University of Southern California Dissertations and Theses

Understanding human-building interactions through perceptual decision-making processes

(USC Thesis Other)

Understanding human-building interactions through perceptual decision-making processes

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content UNDERSTANDING HUMAN-BUILDING INTERACTIONS THROUGH
PERCEPTUAL DECISION-MAKING PROCESSES

by

Gokce Ozcelik

A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(CIVIL ENGINEERING)

Guidance Committee Members:
Dr. Burcin Becerik-Gerber (chair), Dr. Lucio Soibelman, Dr. Gale Lucas

August 2019

I
TABLE OF CONTENTS
Acknowledgement ............................................................................................................................................................ 1
Chapter 1. Executive Summary ..................................................................................................................................... 3
Chapter 2. Motivation ................................................................................................................................................... 11
Chapter 3. Background ................................................................................................................................................. 15
3.1. HUMAN ENERGY CONSUMPTION BEHAVIOR ............................................................................................................................ 15
3.2. DECISION-MAKING .................................................................................................................................................................... 16
3.3. HUMAN PHYSIOLOGICAL, PSYCHOLOGICAL AND PHYSICAL RESPONSE TO BUILDING SYSTEM OUTPUT STIMULI .............. 20
3.3.1. Human response to thermal stimuli ........................................................................................................................................ 21
3.3.2. Human response to visual stimuli ............................................................................................................................................ 22
3.3.3. Human psychological response to thermal and visual stimuli .................................................................................. 23
Chapter 4. Literature Review ..................................................................................................................................... 25
4.1. VIRTUAL ENVIRONMENTS AS A RESEARCH TOOL .................................................................................................................... 26
4.2. THERMOCEPTION IN VIRTUAL ENVIRONMENTS ...................................................................................................................... 27
4.3. THERMAL COMFORT AND SATISFACTION TO UNDERSTAND THERMOCEPTION IN VIRTUAL ENVIRONMENTS .................. 31
4.4. HUMAN-BLIND INTERACTIONS ................................................................................................................................................. 33
4.5. HUMAN-LIGHTING SYSTEM INTERACTIONS ............................................................................................................................. 36
4.6. HUMAN-THERMOSTAT INTERACTIONS .................................................................................................................................... 40
4.7. HUMAN-BUILDING INTERACTIONS IN MULTIMODAL DISCOMFORT CONTEXT...................................................................... 43
4.8. HUMAN-BUILDING INTERACTIONS IN MULTI-OCCUPANCY CONTEXT................................................................................... 47
4.9. ENERGY AND COMFORT IMPLICATIONS OF HUMAN-BUILDING SYSTEM INTERACTIONS ..................................................... 50
4.9.1. Energy and comfort implications of human-blinds and lighting systems interactions ................................ 50
4.9.2. Energy and comfort implications of human- thermostat interactions ................................................................. 51
Chapter 5. Objectives and Research Questions ..................................................................................................... 55
Chapter 6. Benchmarking Thermoception in Virtual Environments to Physical Environments for
Understanding Human-Building Interactions ....................................................................................................... 56
6.1. METHODOLOGY .......................................................................................................................................................................... 58
6.1.1. Design of experiment ...................................................................................................................................................................... 59
6.1.2. Environment and apparatus ....................................................................................................................................................... 60
6.2. EXPERIMENT DETAILS ............................................................................................................................................................... 62
6.2.1. Recruitment ......................................................................................................................................................................................... 62
6.2.2. Pre-experiment session .................................................................................................................................................................. 63
6.2.3. Experiment session ........................................................................................................................................................................... 63
6.3. POST-EXPERIMENT SESSION ..................................................................................................................................................... 66
6.4. RESULTS AND DISCUSSION ........................................................................................................................................................ 67
6.4.1. Actual versus perceived indoor air temperature .............................................................................................................. 68
6.4.2. Thermal comfort and satisfaction............................................................................................................................................ 70
6.4.3. Number and type of interactions .............................................................................................................................................. 72
6.5. LIMITATIONS AND FUTURE STUDIES ........................................................................................................................................ 75
6.6. CONCLUSIONS ............................................................................................................................................................................. 77
Chapter 7. Understanding Human-Building Interactions Under Multimodal Discomfort in Single
Occupancy Offices .......................................................................................................................................................... 79
7.1. DESIGN OF EXPERIMENT ........................................................................................................................................................... 81
7.2. ENVIRONMENT AND APPARATUS .............................................................................................................................................. 82
7.3. EXPERIMENT DETAILS ............................................................................................................................................................... 83
7.3.1. Recruitment ......................................................................................................................................................................................... 83
7.3.2. Pre-experiment session .................................................................................................................................................................. 84
7.3.3. Experiment session ........................................................................................................................................................................... 84

II
7.3.4. Post-experiment session ................................................................................................................................................................ 87
7.3.5. Data analysis ....................................................................................................................................................................................... 87
7.4. RESULTS AND DISCUSSION ........................................................................................................................................................ 88
7.4.1. Perceptual decisions ........................................................................................................................................................................ 89
7.4.1.1. Type of decisions ........................................................................................................................................................................ 89
7.4.1.2. Number of decisions .................................................................................................................................................................. 92
7.4.1.3. Occurrence probabilities of decisions ................................................................................................................................. 93
7.4.1.4. Response time .............................................................................................................................................................................. 95
7.4.1.5. Decision patterns ........................................................................................................................................................................ 96
7.5. LIMITATIONS AND FUTURE STUDIES ...................................................................................................................................... 100
7.6. CONCLUSIONS ........................................................................................................................................................................... 101
Chapter 8. Understanding Human-Building Interactions Under Multimodal Discomfort in Multi
Occupancy Offices ........................................................................................................................................................ 103
8.1. METHODOLOGY ........................................................................................................................................................................ 103
8.1.1. Data analysis .................................................................................................................................................................................... 105
8.2. RESULTS AND DISCUSSION ..................................................................................................................................................... 106
8.2.1. Multi-occupancy survey responses ........................................................................................................................................ 107
8.2.2. Perceptual decisions ..................................................................................................................................................................... 110
8.2.2.1. Type of decisions ...................................................................................................................................................................... 110
8.2.2.2. Kind of decisions ....................................................................................................................................................................... 112
8.2.2.3. Number of decisions ................................................................................................................................................................ 116
8.2.2.4. Response time ............................................................................................................................................................................ 118
8.3. LIMITATIONS AND FUTURE STUDIES ...................................................................................................................................... 120
8.4. CONCLUSIONS ........................................................................................................................................................................... 121
Chapter 9. Understanding the Reasons of Human-Building Interactions .................................................... 124
9.1. METHODOLOGY ........................................................................................................................................................................ 125
9.1.1. Data analysis .................................................................................................................................................................................... 128
9.2. RESULTS AND DISCUSSION ...................................................................................................................................................... 128
9.2.1. Underlying reasons of HBIs....................................................................................................................................................... 129
9.2.2. Reported reasons of HBIs ........................................................................................................................................................... 133
9.3. LIMITATIONS AND FUTURE STUDIES ...................................................................................................................................... 137
9.4. CONCLUSION ............................................................................................................................................................................ 138
Chapter 10. Limitations and Potential Future Work ......................................................................................... 141
10.1. LIMITATIONS .......................................................................................................................................................................... 141
10.2. POTENTIAL FUTURE WORK: OCCUPANT BEHAVIOR MODELLING AND ENERGY CONSEQUENCES OF BEHAVIORS ........ 143
Chapter 11. Conclusions............................................................................................................................................. 145
References ..................................................................................................................................................................... 149

III
TABLE OF FIGURES
FIGURE 1. PLAN OF THE OFFICE SPACE AND LOCATION OF THE EXPERIMENT WORKSTATION .......................................................... 60
FIGURE 2. (A) AND (B) PHYSICAL OFFICE SPACE; (C) AND (D) VIRTUAL OFFICE SPACE .................................................................... 62
FIGURE 3. TYPE OF INTERACTIONS PROVIDED IN VIRTUAL (LEFT) AND PHYSICAL (RIGHT) ENVIRONMENTS [(1) LOCAL
HEATER/COOLER, (2) HOT/COLD BEVERAGES, (3) DESK FAN, (4) AD THERMOSTAT, (5) RADIANT HEATER] .................... 65
FIGURE 4. PERCENT DISTRIBUTIONS OF TYPE OF FIRST INTERACTIONS IN HOT/COLD CONDITIONS ............................................... 74
FIGURE 5. EXPERIMENT OFFICE SPACE (INITIAL CONDITIONS) (TOP: NORTH-FACING OFFICE; BOTTOM: SOUTH-FACING OFFICE)
....................................................................................................................................................................................................... 81
FIGURE 6. A PARTICIPANT IMMERSED IN THE EXPERIMENT SETTING THROUGH OCULUS RIFT HMD AND CONTROLLING THE
ENVIRONMENT THROUGH A CONTROLLER .................................................................................................................................. 84
FIGURE 7. INTERACTION MEANS: (1) DESK FAN, (2) THERMOSTAT, (3) RADIANT HEATER, (4) TASK LAMP, (5) CEILING LAMP,
(6) BLIND ...................................................................................................................................................................................... 85
FIGURE 8. NUMBER OF DECISION-MAKERS VERSUS DECISION NUMBER INDEX (A) THERMAL AND VISUAL, (B) THERMAL, (C)
VISUAL DECISIONS ......................................................................................................................................................................... 93
FIGURE 9. DETAILED DECISION PROBABILITIES IN THE NO DISCOMFORT CONDITION (DIFFERENT TONES OF THE SAME COLOR
REFER TO THE SAME REMEDY) .................................................................................................................................................... 94
FIGURE 10. DETAILED DECISION PROBABILITIES IN THE MULTIMODAL DISCOMFORT DECISION (DIFFERENT TONES OF THE SAME
COLOR REFER TO THE SAME REMEDY) ........................................................................................................................................ 95
FIGURE 11. DECISION PATTERNS TWO CONDITIONS ........................................................................................................................... 97
FIGURE 12. SINGLE (LEFT) AND MULTI-OCCUPANCY (RIGHT) OFFICE ............................................................................................. 104
FIGURE 13. SOCIAL AVOIDANCE AND DISTRESS WITH RESPECT TO DISCOMFORT LEVEL ................................................................ 108
FIGURE 14. METRICS MEASURED IN PRE- AND POST- EXPERIMENT SURVEYS ................................................................................. 126
FIGURE 15. REPORTED REASONS OF INTERACTIONS MEASURED THROUGH POST-EXPERIMENT SURVEYS .................................... 127
FIGURE 16. SANKEY VISUALIZATION OF DECISIONS MAPPED ON REPORTED REASONS IN COMBINED SAMPLE ............................. 135
FIGURE 17. FREQUENCY OF DECISIONS PER REPORTED REASON IN NO DISCOMFORT CONDITION ................................................. 136
FIGURE 18. FREQUENCY OF DECISIONS PER REPORTED REASON IN MULTIMODAL DISCOMFORT CONDITION ............................... 137

IV

TABLE OF TABLES
TABLE 1. FORMS OF STIMULI (REPRESENTED BY TYPE OF ENERGY)-STIMULATED SENSE TABLE [85] ........................................... 21
TABLE 2. PERCENTAGE OF CORRECTLY AND INCORRECTLY (LOWER OR HIGHER THAN IT IS) REPORTED PERCEIVED
TEMPERATURE .............................................................................................................................................................................. 69
TABLE 3. MEAN AND STANDARD DEVIATION FOR THE NUMBER OF INTERACTIONS UNDER BOTH HOT AND COLD CONDITIONS IN
VIRTUAL AND PHYSICAL OFFICES ................................................................................................................................................. 72
TABLE 4. VISUAL AND THERMAL ENVIRONMENT PREFERENCES OF PARTICIPANTS IN TWO CONDITIONS ....................................... 89
TABLE 5. TYPE OF IMMEDIATE DECISIONS IN THE TWO CONDITIONS ................................................................................................. 90
TABLE 6. DECISIONS IN TWO CONDITIONS ........................................................................................................................................... 92
TABLE 7. MEANS AND MEAN COMPARISONS OF IMMEDIATE AND FINAL RESPONSE TIMES AND THE RESPONSE RATE ................... 96
TABLE 8. VISUAL AND THERMAL ENVIRONMENT PREFERENCES OF PARTICIPANTS IN ALL CONDITIONS ....................................... 107
TABLE 9. TYPE OF IMMEDIATE DECISIONS.......................................................................................................................................... 111
TABLE 10. KIND OF DECISIONS ........................................................................................................................................................... 114
TABLE 11. UNDERLYING REASONS SIGNIFICANTLY ASSOCIATED WITH DECISIONS IN MULTI OCCUPANCY OFFICE ....................... 115
TABLE 12. REPORTED REASONS SIGNIFICANTLY ASSOCIATED WITH HBI DECISIONS IN MULTI-OCCUPANCY OFFICE .................. 116
TABLE 13. NUMBER OF DECISIONS ..................................................................................................................................................... 118
TABLE 14. MEANS AND MEAN COMPARISONS OF IMMEDIATE AND FINAL RESPONSE TIMES AND THE RESPONSE RATE .............. 120
TABLE 15. UNDERLYING REASONS SIGNIFICANTLY ASSOCIATED WITH DECISIONS IN COMBINED SAMPLE ................................... 133
TABLE 16. REPORTED REASONS SIGNIFICANTLY ASSOCIATED WITH DECISIONS IN COMBINED SAMPLE ....................................... 134

1
Acknowledgement
Dedicated to my loving parents, Aysema & Ahmet Ozcelik, brother, Goktug Ozcelik, and
grandmother, Mesude Yilmaz-Karsli, for encouraging and supporting me throughout my doctorate
journey.
First and foremost, I would like to express my sincere gratitude to my PhD advisor Dr. Burcin
Becerik-Gerber. It has been a privilege to work with you. I genuinely appreciate your wholehearted
support and guidance throughout my PhD. I admire your success and professionalism as a woman
engineer, professor and researcher. I am very thankful to you for being by my side during my
journey of growing as a researcher. I will cherish the relationship we built throughout my career.
I would like to acknowledge the honorable members of my dissertation committee, Dr. Lucio
Soibelman and Dr. Gale Lucas, for their continuous support and guidance. Thank you for making
this journey remarkable for me. I am truly grateful for the time you took to advise me on my
dissertation studies.
I thank the entire Civil and Environmental Engineering Department at the University of Southern
California and iLab team. I appreciate the faculty, staff, students who worked with me, took classes
with me, and all the friendships we built during my graduate studies.
I thankfully acknowledge the support of National Science Foundation Grant number 1351701. Any
opinions, findings, and conclusions or recommendations expressed in this material are those of the
authors and do not necessarily reflect the views of the National Science Foundation.
I would like to thank my family for unconditionally loving, supporting and encouraging me. Mom
and dad, I wish words could describe how much I love and appreciate you. Thank you for being
with me, in my heart and mind all the time despite the thousands of miles between us. My lovely
brother, Goktug, I wish I could witness all the remarkable moments and achievements you

2
experienced while I was away from home. Thank you for being so mature and supportive. I could
not ask for a better gift from life than you.
I owe special thanks to my lifelong mentors Dr. Engin Karaesmen and Dr. Erhan Karaesmen for
supporting me throughout my undergraduate and graduate studies. You will always have a special
place in my heart.
Lastly, I want to share my appreciation for my friends who are here with me and who are far away
but always make me feel like there is no such thing called distance. Thank you for making my life
and PhD journey special.

3
Chapter 1. Executive Summary
In the United States, people spend almost 90% of their lives indoors. Thus, built environments,
especially buildings are important for the quality of life [1]. In the United States, buildings account
for almost 42% of the energy and almost half of it is consumed by commercial buildings. A major
contributor of this consumption is heating, cooling, ventilation and air conditioning (HVAC) (with
almost 43% share) [2], followed by lighting systems (with almost 19% share) [3]. Not only
building system operations but also occupants’ interactions with the building systems (e.g.,
heating/cooling remedies and lighting fixtures), and occupant-related factors (i.e., preferences,
behavior, habits, and so on) influence building energy consumption [4]. Behavior is any response
of an individual or groups to their environment, habit is non-reflective repetitive behavior, and
preferences are the choices made by the individuals [5]. In addition to understanding the
differences between these terms at the macro level, focusing on the decision-making processes
might provide micro level information about the environmental preferences, habits and behaviors.
Previous studies highlighted the gap between actual and estimated building energy consumption
and identified the occupant behavior as the major contributor of this inaccuracy and pointed out
the strong relation between occupant behavior and energy consumption [6–12]. Uncertainties
regarding occupant behavior [12, 13] and simplified occupant-related assumptions [14, 15] are key
factors limiting accurate building energy consumption predictions that could be obtained by
simulation tools. Although individual interactions are hard to predict, interaction related trends
and patterns for group of building occupants that could be retrieved from observation studies [16]
could potentially provide insight regarding human-building interactions.
There are two major determinants of human behavior: internal (personal factors such as,
individual, physical or mental conditions that influence the human behavior) and external

4
(contextual, physical and system-related factors that may influence the environment) factors [17].
Environmental psychologists who focus on pro-environmental behavior also derived the
determinants of energy consumption behavior from these internal and external factors [18–20]. In
this dissertation, we compiled these determinants by adopting Stern’s ‘Attitude-Behavior-Context
Model’ theory [21] as a framework towards modeling the human-building interactions through
perceptual decision-making. In the ABC theoretical decision-making approaches, (B) stands for
the behavior that is a product personal-sphere attitudinal variables (A) (i.e., personal beliefs,
norms, values and general potential choices of actions in certain ways under certain circumstances)
and contextual factors (C) [21].
Prior to modelling, it is imperative to understand human-building interactions. Decisions take
place in a built environment, thus, understanding how and why occupants respond to
environmental stimuli (i.e., such as temperature and light stimulus) through the environment-
decision-maker interactions would help us understanding human-building interactions. In this
dissertation, we focused on perceptual decision-making processes (defined as the cognitive process
of choosing the preferred options or course of actions based on sensory evidences and perceptions).
For all decision-making processes, information gathering from the environment is the initial and
the common step. Humans gather information from their microclimate (i.e., in our context offices
are the microclimates of occupants) through environmental cues (in this dissertation we focus on
environmental stimuli produced by heating/cooling remedies and lighting systems as cues).
Wickens explained the information processing during decision-making as follows; an event
(signal, stimulus for living organisms, such as multimodal discomfort through glare and solar heat
gain) takes place in the environment (e.g., an office microclimate), human (e.g., occupant)
perceives it and transfers this information into a response (e.g., decision to interact with the

5
available interaction means to respond) [22]. These responses could be physiological,
psychological, or physical (which is mediated by the former two types of responses). In this
dissertation, we mainly focused on the latter as it stands in the nexus of former two and represents
the interactions with provided remedies (i.e., heating/cooling remedies and lighting fixtures).
Therefore, we collected occupant behavior data regarding energy consumption related interactions
(which are also simple adjustments providing rapid ambient changes [23]; adjusting the blinds,
ceiling lamps, task lamps, thermostat, fan and a heater) in the presence of multimodal discomfort
stimuli to understand human-building interactions through perceptual decision-making.
Human-building interactions with regards to heating/cooling remedies and lighting/daylighting
fixtures, and their modeling and simulation (e.g., [24, 25]) were often studied in siloes. In other
words, understanding occupant behavior was explored through focusing on one interaction mean
at a time (e.g., light switch behavior only, its modelling, and simulation [26]). But, in mundane
office contexts, multiple interaction options (e.g., blinds, thermostat), multiple cues (i.e., visual
and thermal) co-exist and they are interrelated. Despite the siloed approaches, one major common
ground of these interactions is the end-users (i.e., occupants). Occupants’ interactions with
thermostats influence energy consumption and indoor air quality (IEQ), and hence occupants’
comfort and satisfaction [27]. Likewise, lighting and daylighting (interactions) influence the
occupants, their physiological responses (e.g., circadian effects and alertness) [28], mood,
satisfaction [28, 29], well-being and productivity [29–31]. Despite different types of stimuli (i.e.,
thermal and visual), there exist commonalities of occupant interactions with heating/cooling
remedies and lighting/daylighting fixtures; occupants need more and “easy to access” control for
the interaction means [32–36] (e.g., occupants prefer local heaters/coolers, and personalized
ventilation [35, 36]). More and easy to access control increase the thermal [37, 38] and visual [33]

6
satisfaction and comfort, and potentially decrease the building energy consumption [33, 39].
Therefore, in our studies, we provided personalized ventilation, a local heater, a thermostat, ceiling
lamps, a task lamp and a blind as interaction means that had easy to access controls (i.e., all were
equidistant from the participants) to the participants for capturing the high-resolution occupant
interactions.
A key limitation in occupant behavior modelling and building performance simulation context is
the lack of information regarding accurate occupant interactions. In this dissertation, we adopted
a holistic approach of understanding occupant behavior through how they perceive, decide and
respond to multimodal stimuli (e.g., co-presence of visual and thermal stimuli as in mundane office
realism) across different settings (e.g., single- and multi-occupancy offices). Understanding how
and why occupants respond to multimodal stimuli in offices could improve the building system
operations (e.g., lighting and HVAC) and human-centered design. Such holistic approach requires
understanding the occupant interactions in the nexus of thermal and visual cues, in both single-
and multi-occupancy settings. Providing a multimodal discomfort (e.g., co-presence of glare and
solar heat gain) context could enable this holistic understanding by providing co-presence of
thermal and visual cues and understanding occupants’ real needs through their responses to these
cues. In this dissertation, we focused on testing this coupled (i.e., multimodal) discomfort scenario,
by which we retrieved detailed occupant behavior information which could potentially contribute
to future occupant behavior modelling and simulation studies, and building energy management
strategies.
Therefore, the motivation of this dissertation is understanding high-resolution human-building
interactions (i.e., we embodied the interactions through decisions and responses, derived metrics
attributed to interactions, such as number, type, kind and occurrence probability of these

7
interactions) through occupants’ perceptual decisions in contextually rich offices (i.e., wherein
multimodal sensory cues and multi-occupancy are possible). We envision responsive, potentially
energy efficient indoor environments that could respond to occupants’ real needs and expectations.
Our unique contributions will be providing the missing high-resolution occupant interactions
across different office contexts (i.e., north- and south-facing offices, single and multi-occupancy
offices). By this way, future studies could also leverage this rich occupant information to develop
realistic energy consumption behavior models (by incorporating internal and external determinants
of behavior using supervised learning methods to mathematically model the occupant behavior),
and we could inform the future human-centered design and renovation decisions. Thus, in this
dissertation, we focused on how and why occupants decide to interact (i.e., respond to) with
heating/cooling remedies and lighting fixtures in single- and multi-occupancy offices wherein
multimodal sensory cues (i.e., visual (glare) and thermal (solar heat gain)) were present. Our
research objectives were to understand the decision-making processes to create an insight on how
occupants behave and respond to multimodal discomfort in single and multi-occupancy settings.
In this regard, investigation of the perceptual decision-making processes is important for a more
prolific and more comfortable responsive built environments.
To address the gap of lacking human-building interaction information we designed and conducted
three experiments with human subjects. We included multiple subjective surveys in our
experiments to accommodate both internal and external determinants of occupants’ energy related
behaviors. Immersive Virtual Environments (IVEs) provide a multimodal sensory experience (i.e.,
visual, haptic, auditory, olfactory, thermal and gustatory or any combination of them to users and
provide an enhanced control over the extraneous variables to [40] and enables experimenters to
conduct controlled experiments. Thus, we used IVEs to create experiment settings (i.e., single-

8
and multi-occupancy offices) and perceptually envelop the participants in controlled experiment
settings.
Our initial step towards understanding human-building interactions was to benchmark
thermoception in virtual environments to physical environments. To provide multimodal stimuli
(i.e., visual and thermal) in virtual environments, we had to develop a hybrid experiment setting
wherein the thermal stimuli exist in physical environment and conveyed through real
heating/cooling remedies while participants could control them in virtual environments through
gaming controllers. However, providing thermal cues in contextually enriched virtual
environments (i.e., fully equipped with heating/cooling and visual remedies’ controls, furnished)
in human-building interactions context have not been validated. Thus, first we had to ensure the
adequacy of using IVEs in thermoception context. We designed and conducted a human subject
experiment to validate the adequacy of using virtual environments in thermoception context, so
we could add heating/cooling remedies to the physical environment in our human-building
interactions studies to enhance the mundane realism through sensory multimodality. We recruited
fifty-six participants to our experiment. Our findings confirmed the adequacy of using immersive
virtual environments in thermoception context. Thus, we concluded that our proposed
methodology of providing thermal cues in physical environments while participants are immersed
in virtual environments in human-building interactions studies is adequate. Chapter 6 summarizes
the experiment and our findings.
As our next step, we investigated the human-building interactions in single-occupancy offices. We
conducted a human subject experiment in virtual environments. We recruited ninety participants
to this experiment. We provided three heating/cooling remedies (i.e., thermostat, desk fan, local
heater) and three lighting fixtures (i.e., ceiling lamp, task lamp, blinds) to participants. We

9
monitored the participants’ interactions in a single-occupancy office. We used statistical analyses
and data visualization methods for deriving inferences from our high-resolution interaction data.
We used number, type (i.e., thermal and visual), kind (e.g., desk fan, blind), hierarchical order,
occurrence probabilities and patterns of interaction decisions as key analysis metrics. Our findings
are summarized in Chapter 7. Then, we investigated the human-building interactions in multi-
occupancy offices. We conducted another human subject experiment and recruited sixty
participants to this experiment. We pursued the same experiment methodology, similar data
analyses methods and metrics as in our previous study. Our findings are summarized in Chapter
8. In Chapter 9, we focused on understanding how and why occupants decide to interact with
heating/cooling remedies and lighting fixtures in certain ways. For this exploration, we performed
statistical analyses on the metrics we measured in subjective experiment surveys to understand
their relations to interaction decisions.
This dissertation is structured as follows: Chapter 2 describes the problem and our motivation to
focus on this problem. Chapter 3 provides background regarding human energy consumption
behavior, decision-making, and human response to building system output stimuli. Chapter 4
describes the current state of art through an extensive literature review that guided us to identify
the problems and methods to address them. Chapter 5 articulates our objectives and related
research questions. In Chapter 6, we focus on benchmarking thermoception in virtual
environments to physical environments for understanding human-building interactions. Chapter
7 explains human-building interactions in the presence of multimodal sensory discomfort in single
occupancy offices. Chapter 8 explains human-building interactions in multi-occupancy offices.
Chapter 9 focuses on the reported and underlying reasons of human-building interaction

10
decisions. Chapter 10 covers the limitations of our studies and potential future work, and Chapter
11 concludes the dissertation.

11
Chapter 2. Motivation
People spend almost 90% of their lives indoors in the U.S. [1]. Thus, built environments,
especially, buildings are important for the quality of life. In the United States, buildings account
for almost 42% of the energy almost half of which is consumed by commercial buildings. A major
contributor to this energy consumption is heating, ventilation and air conditioning (HVAC)
systems (with approximately 43% energy consumption) [2], and followed by lighting systems
(with 19% electricity consumption) [3]. Not only building system operations but also occupants’
interactions with the building systems and occupant-related factors (i.e., occupant preferences,
behavior, habits and so on) influence building energy consumption [4]. During the design phase
of building life cycle, building performance simulations have been used to predict the energy
consumptions [41]. Previous studies highlighted the gap between predicted and actual building
energy consumption. There is a strong relation between occupant behavior and energy
consumption [6–12]. One of the key factors limiting the capabilities of energy simulation tools is
the uncertainties of occupant behavior [13, 42]. Due to their stochastic nature, individual
interactions are hard to predict [11], but interaction related trends and patterns for group of building
occupants could be retrieved through observation studies [16]. In our studies, we used occupants’
‘decision-making processes’ as an enabler to understand human-building interactions. Thus, the
initial step towards helping with high energy consumption shares is to understand the real needs
and expectations of occupants, how and why they decide to interact with buildings. Once the data
regarding this important step is available, inferences from occupant behavior analytics could be
used for informing the future occupant behavior modelling and building performance simulations,
and building operations (e.g., HVAC set points, lighting schedules) and maintenance (i.e.,

12
renovation, retrofit) management decisions in the existing building stock, and could help with
human-centered designs of the new assets.
Behavior is any response of an individual or groups to their environment, habits are non-reflective,
repetitive behaviors while preferences stand for the choices made by humans [5]. In addition to
understanding the differences between these terms at the macro level, focusing on the decision-
making processes may provide micro level information about the reasons for environmental
preferences, behaviors and habits. Obtaining this micro level information, could help building
operators and designers to understand energy consumption behavior and reasons of this behavior
so that it can be used during operation phase and for future user-centered designs (UCD) [5].
However, how occupants actually decide to adjust their environment upon microclimatic changes
(e.g., variations in temperature stimuli in the workplace) and to which responses these decisions
are transformed into have not been studied.
Decisions take place in an environment, thus, understanding the environment-decision-maker
interactions facilitates understanding how and why occupants respond to environmental stimuli
(that are outputs of building systems, such as temperature stimulus and light stimulus). Multiple
decision-making processes have been identified in literature. Some of these processes have been
categorized as classical (normative/rational decision-making process) while some are more recent
and realistic in terms of dynamic complex real-life conditions considerations (descriptive and
naturalistic decision-making processes). For all decision-making processes, information gathering
from the environment is a common step. Humans gather information from their microclimate (i.e.,
offices of commercial buildings are considered as microclimates of occupants in this dissertation)
through environmental cues (focal environmental cues for information gathering in this
dissertation are environmental stimuli produced by heating/cooling remedies and lighting

13
systems). For considering the neural aspects of decision-making, the perceptual decision-making
process has been used by neuroscientists and neurocognitive engineers. Perceptual decision-
making is defined as the cognitive process of choosing the preferred options or course of actions
based on sensory evidences and perceptions [43, 44]. In this dissertation, we use perceptual
decision-making as an enabler to understand human-building interactions.
For improved building systems’ operation and management (i.e., lighting and HVAC systems in
this dissertation) by integrating the real needs and expectations of end-users (i.e., occupants),
decision-making processes of humans in their workspaces (i.e., scope here is commercial office
buildings, scale is for single- and multi-occupancy offices) need to be investigated. There exist
studies focusing on the operation strategies for energy consumption reduction of building HVAC
systems [45–49], lighting systems [26, 50], and coupled HVAC and lighting systems [51–54].
However, there is no holistic study investigating the complex-coupled energy consumption
implications of lighting fixtures and heating/cooling remedies, which requires understanding
human behavior through decision-making processes. This problem has negatively impacted the
accurate estimation of energy consumption in commercial buildings, as well as the comfort and
satisfaction of occupants with their indoor environments. Possible causes for this problem are
implementation of generalized standards for design and operation of buildings, in which system-
system interactions, human-system interactions, and occupants’ actual needs and preferences are
missing.
The motivation of this dissertation is to understand the occupants’ decision-making processes with
regards to energy consuming interactions. Thus, through our experimental studies we investigated
how and why occupants decide responding to coupled environmental lighting-temperature stimuli
(i.e., multimodal stimuli). Focal point of this dissertation is the investigation of decision-making

14
processes resulting in occupants’ responses to the building system output changes. The research
objectives of this dissertation are to understand the decision-making processes to create an insight
on how and why occupants behave and respond to multimodal discomfort in single- and multi-
occupancy settings. In this regard, investigation of the perceptual decision-making processes is an
important venue for a more prolific, more comfortable and (potentially) energy efficient and
adaptive built environments.
To address the gap of lack of information about decision-making processes for determining
human-system interactions in commercial buildings, we performed several experimental studies.
Both human subject experiments and surveys were used for understanding the decision-making
processes of occupants giving rise to their responsive actions to environmental stimuli change.

15
Chapter 3. Background
3.1. Human Energy Consumption Behavior
There are two fundamental determinants of human behavior: external factors and internal factors
[17]. External factors are all the contextual, physical and system-related factors that may influence
the environment. Internal factors are the personal factors that are composed of individual, physical
or mental conditions that influence human behavior [17]. Determinants of energy consumption-
related human behavior (e.g. human-building system interactions) in buildings have also been
derived and branched from internal and external factors by environmental psychologists for
studying pro-environmental behavior [18–20]. Stern identified the determinants of pro-
environmental behavior under four different categories: attitudinal factors, contextual forces,
personal capabilities, habit or routine, which are basically derived attributes of the above-
mentioned two fundamental determinants [20]. He also emphasized that separate consideration of
these factors may be misleading to determine the environmental behavior. In another study, Stern
identified the determinants of human dimension of energy use of individuals, households and
organizations as habits, trust, personal values, word-of-mouth communication, and recent personal
experiences and emphasized that ignoring these behavioral human aspects increase building
energy consumption [55]. A previous study also identified the determinants of building energy
consumption as: climate, building related characteristics, user-related characteristics, building
systems and their operation, building occupants’ behavior and activities, social and economic
factors, indoor environmental quality which are components internal and external factors [4].
ABC Theory, formulated by Guagnano et al., has an integrated approach to combine internal and
external factors for better understanding the role of behavior in organism-environments interaction
and eliminating the misleading evaluation of single-factor analysis of behavior [21]. ABC Theory

16
indicates that behavior (B) is a product of personal-sphere attitudinal variables (A) (i.e. personal
beliefs, norms, values and general potential choices of actions in certain ways under certain
circumstances) and contextual factors (C). The relation between these interdependent determinants
of behavior is summarized as, attitude-behavior interaction is maximum when the contextual
factors are neutral and while it approaches to zero when contextual factors are extremely positive
or negative. In this dissertation, we use ABC Theoretic framework to integrate the internal and
external determinants of occupants’ energy consumption behavior to understand the reasons of
human-building interactions.
3.2. Decision-Making
Human decision-making is defined as the cognitive process of choosing the preferred options or
courses of actions from the available alternatives, on the basis of certain information and
considerations [56, 57]. Classical decision theory focuses on the optimal ‘rational/normative’
decision-making [58]. The fundamental assumption of classical decision theory is that plugging
values (i.e., determinants) of choices identified by researchers into the mathematical models would
yield the optimal choice. They focus on what people should ideally do, not necessarily how people
actually perform the decision-making tasks. Normative decision-making’s focal point is the
concept of ‘utility’ in Subjective Expected Utility Theory [59] by which decision-makers value the
outcomes of their decisions, thus decision-making results in optimizing the utility. Expected Utility
Theory has been widely applied in economics and economic behavior studies and psychology [60–
63]. Despite the alternative model and theory developments, normative decision-making models
fail to formulate how actually decision- makers make decisions since they try to reach an optimum
‘golden standard’ solution [64]. However, studies on decision-making uncertainty showed that
humans use various heuristics for making a choice [65, 66].

17
In order to better understand how people actually perform decision-making tasks, ‘descriptive
decision-making models’ were developed. These descriptive models are beneficial for describing
the decision-making behavior [67]. Previous studies showed that human decisions frequently
deviate from the key assumption of the rational models (i.e. plugging values (i.e., determinants) of
choices identified by researchers into the mathematical models would yield the optimal choice)
[67]. Descriptive decision models are not taking the experience of decision-makers into
consideration [68]. As the most recent approach, ‘naturalistic decision-making’ has been studied
for understanding how people actually make decisions in natural environments (i.e., complex real-
world situations such as modeling human behavior during emergency evacuation [69, 70] and fire
evacuation of buildings based on cue and occupant related factors [22, 67, 69, 71, 72]. Lastly,
neuroscientists have studied, specifically, ‘perceptual decision-making’ for predicting choices
(i.e., preference or goal-oriented selections from a set of alternatives), decisions and actions [73].
Significance of perceptual decision-making is indicated as quoted from Leon’s study: “A
prerequisite for any animal’s survival is the ability to make decisions about its sensory world. For
a monkey to respond to a call signaling imminent danger, or a human to determine when to cross
the street based on the flow of traffic” [44]. With the same analogy, while occupants are fulfilling
their physiological and comfort requirements in their microclimate, their decision-making
processes are triggered by the environmental stimuli and outcomes of human-building system
interactions. Information gathered through sensory systems influence how humans behave in the
world and this process is referred to as perceptual decision-making [43]. Perceptual decision-
making is broadly defined as a cognitive process of choosing the preferred options or course of
actions based on sensory evidences and perceptions [43, 44]. Perceptual decision-making is not
only driven by sensory system but also by attention, task difficulty, prior probability of event

18
occurrence, potential outcomes of the events. The neural architecture of perceptual decision-
making is comprised of four interacting steps. The first step is the accumulation and comparison
of sensory evidence, the second step is the detection of perceptual uncertainties, during the third
step decision variables are identified and lastly the fourth step is for performance monitoring
during which the error detection takes place and decision strategies are adjusted accordingly [43].
Although naturalistic decision-making has been used for modeling human behavior in stochastic
environments, perceptual decision-making is more cogent for leveraging micro-level information
related to sensory cues. Thus, in this dissertation we focus on perceptual decision-making for
understanding human-system interactions based on sensory cues in the presence of multimodal
discomfort (i.e., co-presence of changes and variations in multiple types of stimuli, in our context
thermal and visual stimuli, specifically solar heat gain and glare will be the stimulators) to
understand human-building interactions.
Mathematical models have been developed for evaluating the outcomes of assumed underlying
cognitive processes (decision strategies) of decision-making [74]. Neuroscience and psychology
studies of perceptual decision-making have focused on the response time (RT) and accuracy to
two-choice tasks for investigating the pivotal role of decision-making for transforming perception
and cognition into action [75]. Studies that focused on building energy consumption related
decision-making processes studied pre-occupancy (e.g. decisions made by the stake-holders (e.g.,
contractors, architects, engineers) of buildings [76–78]) and post-occupancy (e.g. operation
decisions made by the occupants or retrofitting decisions [79–81] made by the building owners
during post-occupancy phase). However, occupant responses based on perceived environmental
cues have not been mathematically modeled for quantifying human-building interactions.

19
Understanding how and why occupants actually make decisions to interact with their environment
would be helpful for developing accurate occupant behavior models.
Every decision is made within a decision environment, which is defined as a sequence of collecting
information, alternatives, values, and preferences available within the time of the decision [82].
Thus, for understanding and assessing the fundamental information related to occupant decision-
making in indoor built environments and investigating how occupants interact with the building
systems, environmental cues/perceptions need to be investigated. Wickens proposed a model [22]
for information processing during decision-making. In this model, cues (environmental cues in our
studies) were selected and sampled, hypotheses were generated through retrieval from long-term
memory, possible actions were retrieved from long term memory, and an action was selected on
the basis of risks [76] and values of their outcomes [22]. Wickens defined human operators (i.e.,
occupants in the scope of this dissertation) as transmitters of information [22] and summarized the
process as follows; an event (signal, stimulus for living organisms) takes place in the environment,
human perceives it and transfers this information into a response. Wickens emphasized that this
procedure is applicable to any situation where human operators perceive any environmental
change and respond to the change through transmission of the environment. Thus, we could
conclude, changes or perturbations in the microclimate of an indoor environment (i.e. events,
stimuli) are being perceived by the occupants of the environment (i.e. perception), information
transmission results in response of the occupants (i.e. response/interaction with the environment).
In the built environment context, response could be through Human-Building-Interactions (HBIs)
(i.e., bi-directional interactions between built environments and their occupants). For instance, in
response to any micro-climatic change (e.g., increase in ambient room temperature) in an office,
occupants of the office may interact with the heating/cooling remedies to bring the ambient

20
temperature to their comfort levels. In our studies, we collected occupants’ decision (i.e.,
interaction) data in response to indoor environmental changes, specifically lighting and thermally-
related changes to study how and why they decide to interact with the building systems.
3.3. Human Physiological, Psychological and Physical Response to Building System
Output Stimuli
During the process of perception, sensing captures the environmental information (i.e., stimuli).
The strong enough stimulus is transmitted to the brain and cognitive processes in the brain result
in a physical response. Peripheral nervous system (PNS) carries the external signals that respond
to light, touch, sound, temperature and chemicals from external receptors that provide information
about human-environment interaction, and internal receptors that respond to the internal changes
inside the body. PNS transmits this information to the central nervous system (CNS) [83]. Senses
and stimuli classification are shown in the table below [84]. In our studies, we focused on thermal
and visual (i.e., light) stimuli. To investigate how occupants interact with building systems when
they perceive a change in their microclimate through one or both of these stimuli, one needs to
assess how occupants perceive these stimuli physically and how they respond physiologically and
psychologically. Following table summarizes the forms of stimuli, and we focus on
electromagnetic (i.e. visual) and thermal stimuli in this dissertation. The reason for us to focus on
these two stimuli is the large energy consumption share of HVAC systems, followed by lighting
systems. Providing the co-presence of these two stimuli could help us to understand the occupants’
interactions relatable to buildings’ thermal and visual systems operations.

21
STIMULI SENSE
ELECTROMAGNETIC Vision
MECHANICAL Hearing
Touch
Pain
Vestibular
Kinesthetic
THERMAL Cold
Hot
CHEMICAL Taste
Smell
Table 1. Forms of stimuli (represented by type of energy)-stimulated sense table [85]
3.3.1. Human response to thermal stimuli
There exist four physical determinants of a thermal environment; air (or water) temperature,
humidity, air (or water) movement, and temperature of body surfaces [86, 87]. Thermal stimuli
can be sensed as cold or hot [85], influence of these sensations on human response will be
discussed in the following sections. In cold environments, humans’ decision of staying in cold or
changing the cold environment has been indicated as subjective assessment of perceived skin and
core temperature. Failing to make the right decision for maintaining the thermal balance of the
body may result in dangerous situations. If skin temperature drops below 35.5°C, the sensation of
cold increases, while the maximum cold sensation has been reported to take place at 20°C skin
temperature. Pointing out the problem of subjective assessment of cold sensation [88], Kroemer et
al. emphasized the significance of measuring ambient temperature, humidity, air movement,
exposure time, and humans’ reactions to these physical measures for better understanding
sensation of cold [83]. There are two types of human responses to cold environment; psychological
actions that are often in the form of psychological adaptations and tolerance, and physiological
actions that take place in the body (e.g. redistribution of blood and increased metabolic heat
production). Physical actions (such as increasing the level of clothing, looking for shelter, and

22
using heaters or other external sources for warming up) that are mediated by physiological and
psychological actions also take place. In hot environments, a human body tries to dissipate the heat
gain to the environment, and the same types of responses are observed with different means.
Humans demonstrate psychological response (by tolerating the warmth), physical response (e.g.
decreasing the level of clothing) and physiological response (e.g. redistribution of blood and
increased metabolic heat production). Overheating may result in dangerous heat stress disorders
such as transient heat fatigue, heat rash, fainting, heat cramps, heat exhaustion and heat stroke
[83].
In our studies, we focused on the occupants’ responses to change in ambient room temperature
and light stimuli in terms of their interactions with the heating/cooling remedies and
lighting/daylighting systems. By this way, we were able to understand the occupants’ decision-
making processes through their responses to indoor environmental stimuli change. Doing so, we
helped unlocking the subjectivity of human sensation and future studies could investigate and
quantify the energy consumption consequences of human behavior based on our findings.
3.3.2. Human response to visual stimuli
Humans’ response to light is assessed in terms of two physical determinants of light; color and
intensity [89]. Human response to lighting color is explained through general human cognitive
response to color of the environment. Birren [90] analyzed color as a system comprised of two
subsystems; warm (i.e. between red and yellow) and cool (i.e. between green and violet). Warm
color perception was associated with blood pressure, pulse rate, respiration, perspiration
acceleration and brainwaves increase. Moreover, muscular reaction and increase in eye blinking
were observed. Contrary to the warm colors, cool colors were associated with relaxation and
slower body processes. Goldstein (1939) showed the difference between warm and cool colors in

23
terms of their influence on human cognition and response. In the experiments performed in blue,
yellow, yellow-green, orange painted rooms, up to 12 points difference in standardized intelligence
tests was observed [91]. Followed by Birren and Golstein’s biologic findings, Chan focused on the
perception of color and materials by office occupants [92]. Chan found (marginally) statistically
significant occupant response to red and blue colors and supported this finding by Birren’s study
that showed reactions to color does not take a long duration. Yasukouchi and Ishibashi emphasized
that color temperature of fluorescent lamps has physiological impacts (i.e., cortical arousal level,
heart rate, blood pressure, body temperature and sleeping) on humans [93]. Another study also
investigated the influence of lighting color temperature on human cortical arousal level with three
color temperature levels; 3000K, 5000K, and 7500K, and found that daylight fluorescent lamps
(i.e., 7500 K) increased the arousal level [94]. Another study found that 7500K is more influential
on mental activity level [95]. The same study also showed that lighting color temperature
influenced body temperature [95]. A previous study showed that light intensity stimulates both
circadian and acute physiological human response [96]. Other studies showed that melatonin
suppression as a circadian response is light intensity dependent [97, 98]. Physical response to these
stimuli is being mediated by physiological [89] and psychological response and it takes place as
human-lighting/daylighting systems in the environment (e.g. switching on/off or dimming
lighting, adjusting blinds, and etc.).
3.3.3. Human psychological response to thermal and visual stimuli
In addition to physiological response, studies showed that occupants may perform psychological
adaptive response to indoor environmental changes. A previous study underlined that when
physiological adaptations are not enough to cope with discomfort created by environmental
stimuli, psychological response helps to compensate for the discomfort. In this study,

24
psychological response was in the form of psychological adaptation based on thermal experiences
and expectations [99]. Aligned with these findings, another study identified the psychological
responses as ignoring or tolerating the source of discomfort [100]. Another study highlighted that
high levels of tolerance could be maintained with increased visual satisfaction with the
environment [101]. Heerwagen and Diamond also identified psychological coping through
ignoring the source of discomfort or concentrating harder on work as psychological responses.
They highlighted that psychological coping takes place if environmental adjustments are not
feasible, other actions are not sufficient to decrease discomfort or cost of action is too high (such
as in the social contexts where other people are present) [23]. However, a previous study showed
that these psychological responses have health and productivity implications [23, 102]. In this
dissertation, we focus on physical responses (i.e., interactions) in the nexus of physiological and
psychological responses. Since the psychological responses might have overseen health and
productivity implications, we also had subjective reasoning questions to identify the psychological
responses if there exists any, and we reported the statistically significant responses.

25
Chapter 4. Literature Review
Behavior is defined as occupants’ actions that influence their indoor environment [11]. During the
design phase of building life cycle, energy simulations have been used to predict energy
consumption of buildings with respect to design [41]. Previous studies [11, 12] highlighted the
significant gap between the predicted energy performance of buildings and their measured actual
performance. There is a strong relation between human behavior and energy consumption [6–12].
Uncertainties regarding the behavior of building occupants are one of the key factors limiting the
ability of energy simulation tools to accurately predict building energy requirements [13, 42].
Moreover, as a part of occupant behavior, individual interactions are hard to predict [11], but
interaction related trends and patterns for group of building occupants could be retrieved through
observation studies [16]. In this dissertation, we use ‘decision-making processes’ as enablers to
understand occupant behavior for quantifying the commercial building energy consumption as a
result of human-building systems interactions that take place during the response phase of
decision-making. Understanding how and why human operators (i.e., occupants) interact with
building systems are crucial for developing behavior models to be used in simulations for
understanding comfort and energy consequences of these interactions [12, 103].
HVAC and lighting systems are the foremost two energy consumers in commercial buildings. In
this dissertation, we focus on thermal and lighting stimuli for understanding energy consuming
human behavior through human decision-making. One of the most challenging part of considering
these systems and cues as energy consumers is that they are interrelated. Opening the blinds, or
windows could result in increased or decreased heat load while they influence amount of light
indoors. Previous studies showed that occupants perform adaptive actions in response to the
changes in their physical environment through adjusting their environments (where human-system

26
interactions take place) and through personal changes (e.g., adjusting clothing level, drinks and so
on) [104, 105]. A previous study referred to the occupants’ response as ‘coping behavior’ and
found that people mostly changed ambient conditions with simple adjustments (e.g.,
opening/closing the blinds, turning lights on/off, adding lamps, fans, heaters) that provide rapid
changes in the environment. This study also highlighted the increased personal comfort with
adding fans, heaters, lamps, and found that people also show coping behaviors as drinking a
beverage, going outdoors, walking around, talking to coworkers, and psychological coping.
However, they identified the latter as ‘not a solution’ and the others as slow solutions [23]. We
reviewed human-system interactions in three categories, through which occupant decision-making
and response to stimuli of these systems, are analyzed: human-blinds interaction, human-lighting
systems interaction and lastly, human heating/cooling remedies (i.e., thermostat, fan, radiant
heater) interactions. Extending the number of human-system interactions to these three, we aimed
to investigate the coupled human-system interactions, in addition to the single interactions.
4.1. Virtual Environments as a Research Tool
Immersive Virtual Environments (IVEs) provide a multimodal sensory (i.e., visual, haptic,
auditory, olfactory, thermal and gustatory, and any combination of them [40] experience to users,
wherein they are perceptually enveloped (i.e., immersed) by continuous stream of stimuli and
when users are enabled to interact with the environment [106]. IVEs facilitate an increased level
of perceived realism without decreasing the experimental control, and also provide an enhanced
control over the extraneous variables, while experimental studies in physical environments
potentially require resources (e.g., time and cost), and might have lower experimental control [40].
IVEs have been used in numerous studies in wide-range of fields (e.g., psychology, visual
perception, education and training, psychotherapy, social psychology [40]). Despite the wide use

27
and acceptance of IVEs, and their ability to handle multimodal sensory cues, they have not been
used for understanding human-building interactions and perceived thermal comfort in built
environments. In fact, there are few studies integrating the thermal stimuli to IVEs for enhancing
the immersion experience [107, 108]. To leverage the use of IVEs in thermoception context,
especially in human-building interactions studies wherein humans are the foreground and the main
perceptors, virtual environments need to be benchmarked to physical environments. Due to
enhanced visual power of virtual environments, previous studies used virtual environments in
visual comfort in built environments context [109, 110]. For doing so, they also highlighted the
necessity of testing the adequacy of using IVEs with regards to visual stimuli (i.e., lighting in built
environments), and showed that virtual environments could be used with regards to visual stimuli
[111].
4.2. Thermoception in Virtual Environments
Virtual environments provide experimental control and supremacy in mundane realism and
continuous stimuli. They have been used in multiple disciplines where humans are in the
foreground. To ensure the use of virtual environments for understanding occupant behavior,
interactive options must be adequate representations of reality. Previous studies benchmarked
virtual environments to physical environments with regards to several parameters (e.g., occupant
task performance [111], occupant visual perception [112]). In terms of thermal stimuli, there are
studies that integrated thermal cues in virtual environments, focusing on the local thermoception
of the human body rather than thermal comfort or satisfaction. These studies are mostly based on
the thermal contact phenomena by improving the haptics in virtual environments, which can
potentially ameliorate human computer interactions by creating tactile thermal stimuli on human
skin. Users can touch and grasp objects in virtual environments and yet they still cannot perceive

28
them completely due to the missing cues, such as their temperature. Demi et al. used a thermally
actuated data glove for simulating thermal stimuli on hand. They benchmarked the thermoception
of virtual objects to real objects. Their findings were supportive for providing the missing haptic
component [113]. In another study, Bergamasco et al. focused on the local thermoception to model
thermal feedback in virtual environments. For this purpose, they modeled a finger-object system
(i.e., thermal modeling of finger, and modeling of temperature distribution with contact) in virtual
environments [114–120]. Other studies also focused on hand-object interactions through
thermoception to improve the immersion in virtual environments by developing haptic
displays/devices. These thermo-tactile haptic studies leveraged the force and temperature factors
on the skin to facilitate the object and material identification in virtual environments for improved
immersion.
Several other studies also investigated different ways to improve immersive thermal experiences.
For example, Peiris et al. integrated thermal haptic feedback to the head-mounted display (i.e.,
HMD). They provided cold and hot stimuli to the face through peltiers attached on the HMD for
measuring the directional cue recognition and immersive thermal experience [121]. Ranasinghe et
al. also focused on integrating thermal (i.e., ambient temperature sensations through peltier
modules) and wind (i.e., through small fan modules) stimuli to the HMDs. They showed improved
sensory realism of virtual environments and provided a better sense of presence with this
integration. The authors highlighted the importance of integrating thermal cues to contextually
rich, interactive environments [122]. In addition to the thermo-haptic studies, a previous study
qualitatively showed the influence of climate (i.e., season, daytime, and temperature) on the
perception and use of urban spaces through virtual reality. However, this study focused on the role
of visual cues in climate perception rather than focusing on thermoception in virtual environments

29
[123]. Another application of local thermoception is thermal pain. Humans perceive temperatures
under 15 Celsius degrees and above 35 Celsius degrees as thermal pain [114]. In this context,
previous studies showed that both Virtual Reality (VR) [124–126] and Augmented Reality (AR)
[127] distracted the patients diagnosed with burn injury and alleviated the perception of thermal
pain. Although [128] highlighted the importance of incorporating environmental factors (e.g.,
ambient temperature, clothing) in virtual environments to improve user interactions regarding
thermoception, they chose studying only the effect of clothing on thermal feedback perception.
Studies in the computer graphics domain [108, 129] added the temperature component to the VR
systems. Dionisio emphasized the importance of considering human senses to improve the
immersion experience in virtual environments, as well as to improve the haptic aspects of VR
tools. This study was an important attempt to link the VR graphics interface with heat perception.
Temperature was defined as an outstanding new parameter to bring novelty to the VR interfaces
and presence experience. Temperature stimulation was provided through a ventilator, infrared
lamps, and peltier elements for local skin stimulations [108]. Likewise, Dionisio introduced this
temperature stimulation system as a VR Thermal Kit [129]. This study concluded that temperature
perception is possible using such a kit and could be implemented to normal comfort temperatures
as in office spaces. Later on, Hulsmann built up on the same idea of improving the presence by
including human senses in VR. However, the focus was on the wind simulation in a CAVE (Cave
Automatic Virtual Environment), with results showing that inclusion of wind improved the user
presence in the virtual environment [107, 130]. Frohlich and Wachsmuth added wind and warmth
to the CAVE to combine multi-sensory stimuli in virtual environments [131]. In another study, the
same authors provided wind stimuli as a thermal cue in addition to the visual, auditory, and haptic
modalities. Their results showed that the full combination of these sensory modalities improved

30
the presence in virtual environments while a similar influence was not observed when these
modalities added to the system one at a time. They also suggested adding heat to the system in
future studies [132]. Other studies also added physical wind simulations to VRs to improve the
user experience [133–135]. [133] concluded that more studies are needed to validate the improved
presence, while [135] observed improved subjective presence, and [134] concluded that their
system initiated a venue to integrate wind simulation to virtual environments.
Even though these studies are important steps towards better understanding the human
thermoception in virtual contexts, a thorough understanding of thermoception in virtual
environments, in which continuous thermal stimuli are present, is not well studied. A previous
study on understanding thermal comfort in the context of ubiquitous computing also highlighted
that thermal comfort has not been widely studied by the human computer interaction (HCI)
community either [136]. The majority of the previously mentioned studies focused on local
thermoception (e.g., thermal contact) in virtual environments, and the distractive influence of
virtual environments (e.g., perception of pain related to burns) on thermoception. Closer
approaches that included ambient factors in virtual reality provided wind simulations in a virtual
environment. A recent study included temperature stimuli and measured occupants’ physiological
responses (i.e., heart rate, skin temperature) to thermal changes in an environmental chamber [137]
and compared these responses to the occupants’ physiological responses in a physical
environment. However, none touched upon the use of virtual environments in the context of
thermal comfort in built environments or elaborated on occupants’ physical responses (i.e., human-
building interactions) as response to thermal discomfort.

31
4.3. Thermal Comfort and Satisfaction to Understand Thermoception in Virtual
Environments
Thermal comfort influences overall occupant satisfaction, well-being and performance in built
environments [138]. Humans regulate their body temperature through physiological responses,
which are autonomous responses mediated by the sympathetic nervous system, as well as
behavioral responses through coordinated and voluntary motor activities [139]. Cabanac defined
the ‘thermoregulatory behavior’ as the control of heat gain/loss by adjusting the thermal
characteristics of a physical environment through different means [140]. The motivation of these
responses is the subjective feeling of satisfaction with a thermal environment [140, 141]. During
the process of perception, our senses capture environmental information (e.g., temperature
stimuli). A strong enough stimulus transmitted to the brain results in a physical response, which is
called ‘behavioral temperature regulations’ by Candas and Defour [89]. Under cold and hot
conditions, physiological (e.g., redistribution of blood and increased/decreased metabolic heat
production [142]) and psychological responses (e.g., tolerance of existing thermal condition and/or
adaptation to such condition [143]) take place [83]. These responses mediate adaptive physical
responses [144] through the above-mentioned neurophysiological perception path. Thus, it is
important to study occupants’ thermally adaptive physical responses, such as adjusting fans and
heaters, thermostats, opening/closing windows and doors [35, 145]. A previous study also
highlighted the increased recognition of occupant interactions with building systems as an
important determinant of thermal comfort and satisfaction. Thermal sensation and comfort have
been frequently measured through subjective thermal votes (e.g., ANSI/ASHRAE, Bedford)
derived from different thermal sensation models (e.g., predicted mean vote (PMV), dynamic
thermal sensation (DTS)) [146–149]. However, these models and sensation scales vary in many

32
aspects [150]. A previous study reported that most people do not perceive the categories of these
subjective scales as equidistant. Thus, they concluded that the use of these subjective scales alone
is not sufficient to understand human thermal comfort and sensation. The authors recommended
the use of multi-dimensional measurement methods, such as objective measurements including
physiological and behavioral recordings with subjective votes to understand occupant thermal
comfort [151]. Glicksman and Taub also concluded that comfort models could be improved by
better understanding occupant interactions and comfort state [152]. Thus, in addition to using the
subjective votes, we also used objective metrics (e.g., number and type of adaptive interactions).
Occupants feel more satisfied when they have more control over heating-cooling systems [153–
156]. Previous studies compared personalized conditioning systems (e.g., personal control over
local convective and radiant heating/cooling remedies) to conventional cooling systems (e.g.,
centrally controlled HVAC systems) and showed equal or better thermal comfort with personalized
system [146, 157–161]. In addition, a previous experimental study confirmed that local thermal
stimulation could influence the global thermal perception [147]. Thus, in our studies, we provided
a mixed experience of personalized and conventional heating/cooling for thermal comfort. There
are various comfort determinants, which increase the complexity of thermal comfort [102, 148–
151, 162, 163]. Although the first steady state thermal comfort model (i.e., Fanger’s model) [164]
is commonly used as a thermal environment design criteria and an objective measure in
experimental studies [165], an increasing number of thermal comfort field studies (in physical
environments) show that it is not always a good predictor of actual thermal sensation [166] as it
does not address the personal thermal preferences and related decisions (as in physical
interactions) [11]. Existing models’ (e.g., [164]) limitations to capture and integrate the individual
differences resulted in inaccurate estimation of occupant comfort and led to more human-centered

33
modeling approaches [167–169], permitting direct measurements of perceptions (e.g., [113, 170,
171]). Thus, we pursued a human-centered approach for benchmarking thermoception in virtual
environments to physical environments through direct measurement of perceptions, and their
adaptive consequences (i.e., decision such as physical response to environments through adaptive
interactions).
4.4. Human-Blind Interactions
Building façade is the interaction surface between an indoor environment and an outdoor
environment. Shading systems are one of the interaction components of building facade between
occupants and daylighting. Position and operation patterns of shading systems influence energy
consumption in buildings. However, there exists no agreement on how occupants operate blinds
and what are the underlying reasons of their operation behavior [172].
Studies showed that generally occupants prefer daylighting. However, studies also showed that
occupants’ blind occlusion behavior is for preventing the direct sunlight penetration (e.g., glare
and daylighting), thermal radiation (e.g., overheating) or both [173–175]. Blind interaction or
occlusion behavior through which we analyze human-blind interactions has also two fundamental
determinants; external and internal. External determinants are the ones related to the physical
environment of occupants. External determinants, driving adaptive blind interaction of occupants,
are indoor temperature, indoor lighting, transmitted solar radiation [104], orientation of the
windows, view [172], temporal factors (e.g. time of the day, time of the year), weather conditions,
workstation position, electric lighting characteristic in the room [173]. Searching for the
determinants of blind occlusion patterns, Van Den Wymelenberg concluded that most important
factors influencing blind occlusion are orientation of the building, season, and sky conditions
which are all external determinants [176]. Other studies also identified the variation of blind

34
occlusion motivated by these external factors and they found that the maximum blind occlusion
takes place on the south face [173, 177, 178]. Mahdavi et al. conducted a control-oriented case
study, during which they monitored occupancy, indoor and outdoor temperature, relative humidity,
internal illuminance, external air velocity and global irradiance, status of electrical light fixture
and position of shades. They concluded that lighting and shading system control behavior tend to
be dependent on both internal and external factors [179].
Although there is substantial amount of work in the literature for occupants’ blind use behavior
motivated by the external factors, most of these studies have the limitation of not incorporating the
internal factors through which human-blind interactions could be better investigated by the
involvement of psychological, physiological and other personal factors. Internal determinants of
occupants’ adaptive blind interaction are social, economic, psychological [100], occupant
preferences, habits [173], physiological factors [180]. Day et al. showed that available indoor
lighting influence occupants’ mood, satisfaction, productivity and well-being and found that
occupants open the blinds due to psychological factors and view while they close it due to
physiological reasons [177]. Diversity of these determinants increase the complexity of assessment
of how and why occupants interact with blinds. Inkarojrit coupled some of the external (i.e.,
average luminance of the window, maximum luminance of the window, vertical solar radiation)
and internal predictors (i.e., self-reported sensitivity to brightness) in a case study for investigating
how and why occupants control window blinds in private offices. By coupling these external and
internal factors Inkarojrit envisioned automated intelligent façade systems [180]. However,
previous studies showed that interaction with blinds may also vary depending on the automation
level of the blind systems. Overriding the system through shading system interactions is a common
behavior observed in fully automated lighting systems [181–184]. Previous studies also found that

35
occupants often do not interact with their blinds for a long time (e.g., weekly, monthly or seasonal
interactions were observed) unless there exist discomfort situations due to long term sunlight and
overheating [185]. Another study confirmed this finding; and characterized human-blinds
interactions through passive and active users. First group of passive users tend to keep the blinds
open at all times; the second group tend to keep the blinds closed at all times. Active users open
the blinds upon arrival and close them with increased discomfort (i.e., solar heat gain higher than
50 W/m2) [186]. These findings are consistent with Reinhart’s findings [187], and Sutter et al’s
findings [188]. Sanati and Utzinger found through the surveys they conducted that the major reason
for infrequent interaction with the blinds is their being hard to control [189]. A previous study
revisited human-blind interactions for developing a new algorithm for manual blind switch
behavior. This study highlighted the lower blind-occlusion with better window view, lower blind-
occlusion in private offices compared to open plan offices. The study also identified less often
blind interactions at ground floors than upper floors potentially due to privacy concerns [190].
Limitations:
➢ A previous study noted that there is a need for studying integrated façade/lighting systems
that could operate in all environmental conditions, and their impacts on subjective needs
of occupants and objective building performance [191]. Agreeing with this limitation, we
identified the major limitation regarding human-blinds interactions as the missing holistic
understanding (by integrating the internal and external determinants of occupant behavior)
of how and why occupants interact with the blinds. Once this understanding is articulated,
findings could potentially help with the other needs identified in the previous studies (e.g.,
design and operation of dynamic facades).

36
Studies explained how and why occupants interact with blinds and shading systems by selecting
limited number of parameters of external and internal determinants. Bennet et al. also emphasized
that there exists no consensus on blind use behavior of occupants, hence there is an urging need
for the development of models reflecting occupants’ blind use behavior so that simulations can
better estimate building performance [192]. Although surveyed literature has studies addressing
determinants of blind-use behavior and some studies for understanding why people demonstrate
blind occlusion behavior, there is a gap of empirical studies for investigating the incorporated
internal and external determinants through occupant decision-making mechanisms for
representing and understanding how and why occupants interact with blinds in certain ways.
Moreover, understanding and coupling human perception of environmental cues with the reasons
of their actions (i.e., response) needs to be incorporated thoroughly. Absence of consensus on blind
occlusion behavior (that was highlighted throughout the literature) could be addressed through
empirical studies using decision-making processes.
4.5. Human-Lighting System Interactions
Lighting-related energy consumption makes up of approximately 19% of the total annual energy
consumption in commercial buildings [3]. Human-lighting system interactions could be analyzed
through occupants’ interactions with daylight-related building elements/systems (i.e., blinds) and
artificial lighting systems (i.e., ceiling lamp, task lamp). A building façade is an interface for
daylight and heat transfer between indoor and outdoor environments. In the previous section, we
surveyed human-blind interactions for addressing how and why occupants interact with daylight
systems, specifically shading systems (i.e., blinds). In this section, we focus on how and why
occupants interact with artificial lighting systems. We analyzed human-lighting system
interactions by categorizing the fundamental determinants as external and internal.

37
Previous studies on internal determinants of human-lighting interactions focused on the human
dimension of lighting systems. Previous studies showed the influence of lighting on physiological
response (e.g., ocular light, circadian effects, mental health, and alertness [28]). Previous studies
demonstrated the strong relationship between lighting and occupants’ mood, satisfaction [28, 29]
with their indoor environment, well-being and productivity [29–31]. Focusing on the biological
aspects of lighting rather than physical aspects, Knoop observed that lighting can also increase
alertness, decrease sleepiness and stress so that it can increase well-being and performance [30].
Studies on external determinants of human-lighting interactions focused on; contrast, glare, color
temperature and intensity of lighting [28], context [193], design of lighting [31], lighting controls
[194]. A previous study highlighted the illuminance level as a triggering parameter of lighting
switch on-off behavior [195]. A study on current and potential everyday lighting system controls
for facilitating the design of lighting control interfaces depicted the user interaction experience
through an empirical experiment. In this study, Offermans et al. investigated the generalized design
principles for lighting control domain specific interactions. They identified the determinants of
user (i.e., occupant) interaction with the lighting systems as lighting needs of the users and context.
Context parameters used for defining the requirements for interactions were identified as
environmental conditions, users’ task and intentions, social environments [193]. Within contexts,
human-lighting system dependency on routines (e.g., switching on or off the lights upon arrival)
and activities (e.g., cooking, reading books) were observed. Considering the biological aspects of
lighting for increasing human well-being, Knoop also highlighted the significance of design and
use of lighting depending on the contexts, which Knoop called ‘dynamic lighting’ [30].
Occupant’s adaptive actions through interactions with the lighting systems observed upon change
of environments (e.g., different office rooms) or when drastic change in amount of natural light

38
takes place. Studies showed that occupants’ adaptive switch on/off behavior usually
(approximately 91% of the time) takes place upon arrival or just before departure [32, 196], which
has been associated with balancing the lighting conditions between indoors and outdoors [29, 197].
Likewise, Lindelof and Morel also had similar findings regarding switch on/off behavior.
However, they concluded that this was potentially due to the traditional placement of controls next
to the door instead of close to the desks [103]. Contrary to adaptive lighting behavior of occupants,
in some cases occupants were observed not to use overhead lighting at all during the day [32].
A previous study focused on human-building interactions with regards to window opening and
switch-on behavior [195]. Specifically, the results on switch-on behavior were in accordance with
the previous studies’ findings. This study highlighted the potential energy savings and comfort
improvements through integrating human-building interactions to building performance
simulations (BPS) for more accurate results. While categorizing the occupants as active, medium
and passive, they identified the need for granular probability of occurrence information with
regards to occupant interactions. In another study, Fabi et. Al. also categorized occupants’ as active
and passive with regards to switch on/off behavior. They mentioned that an occupants’ being active
not necessarily means energy savings since they might actively be switching off but also switching
on the lights. They focused on two environmental factors to derive switch on/off probabilities; (1)
room temperature, (2) sun elevation. Having said that, they explicitly highlighted the need for
expanding the environmental conditions for better understanding of human-lighting systems
interactions [198].
Lindelof showed that actual occupant interactions with lighting controls varied a lot between 30
participants in different offices some of which were depicted as active users while some were
passive [103]. Variation of occupant-lighting system interactions need to be supported by the

39
causalities and reasoning behind their interaction means for developing a realistic behavior model
to be used in building energy context. Offermans et al. investigated why or why not occupants
perform adaptive adjustments on lighting systems. Some of the findings identified the causalities
as difficulty of adjustment and insufficient improvement of existing lighting scene [193]. Previous
studies showed that occupants want more control over the lighting systems in their environments
[32, 34]. Wyon et al. also emphasized that user-centered control is more crucial for health, comfort,
and productivity than optimizing the pre-set requirements [27]. Another human-lighting system
interaction type takes place under the existence of lighting systems automation. Despite the
variation of occupant experience with automated lighting systems, previous studies showed that
given the fully-automated environment, occupants still desire control and override the automation
system through their system control interactions. Another study also confirmed the variations in
occupants’ lighting preferences and need for “easy to access” controls in a pilot human-lighting
and blinds interactions study. They observed higher daylight utilization with easy-to-access
controls and anticipated lower energy consumption [33]. Observing the interactions of office
workers with lighting systems in a field study, Escuyer concluded that semi-automatic dimmable
lighting systems (with manual illuminance control), optional task lighting and user-friendly control
interface are the fundamental expectations of occupants from lighting system operations [194].
Limitations:
➢ Studies reviewed in this section lead us to conclude that there exists a need for integrating
ease of lighting control, having multiple control options and/or lighting fixtures (e.g., task
lamps), having user-friendly control interfaces for elaborating understanding human-
lighting system interactions. Thus, in our studies, we ensured the ease of access to the

40
lighting controls and providing multiple lighting fixtures to map the high-resolution
occupant-lighting interactions.
➢ Occupants’ not showing adaptive actions to an indoor climate change does not necessarily
mean they do not want to respond to these changes. Analyzing human-lighting interactions
without assessing the reasons behind their interactions (i.e. how and why they do interact)
could be misleading and result in misrepresentation of human behavior once they are
modeled for energy and comfort quantification.
➢ Existing studies on adaptive behavior are limited to determination of certain behavior types
(e.g. adaptive switch on/off behavior) while the reasoning behind occupant actions were
not assessed in detail.
Use of lighting related behavior as an adaptive measure needs to be coupled with other adaptive
measures (e.g., other perceived environmental cues like temperature) and determinants of behavior
for better understanding how and why occupants interact with lighting systems.
4.6. Human-Thermostat Interactions
Occupants’ interactions with thermostats influence energy consumption and Indoor Environmental
Quality (IEQ) [27]. As indicated in the previous sections, occupant behavior is a key factor of
inaccuracy on building energy consumption including heating-cooling system energy
consumption. Occupants’ thermally adaptive behaviors include adjusting fans and heaters,
thermostats, opening/closing windows and doors when they do not feel thermally comfortable [35,
145]. Modelling occupant-thermostat interactions would contribute addressing the gap between
actual and estimated energy consumption [13]. For modelling comfort-driven occupant behavior,
Langevin et al. used perceptual control theory (PCT) in which occupants interact with their

41
environments based on some reference level perceptions (in this case personal thermal sensations,
and these are related to internal determinants). Behavior is recognized as a by-product of a negative
feedback loop, through which occupants try to control their current perceived environments. Some
of their findings are as follows; personal characteristics (e.g., personal thermal acceptability range)
were found to be one of the determinants (i.e., internal determinants) of comfort and behavior
outcomes and occupants were observed to choose the easiest mean of adaptive actions for setting
the environment to their personal comfort levels (e.g., adjusting the local heaters or fans first, then
interacting with thermostats if still needed) [35]. Although their findings regarding human-thermal
environment interactions are helpful for the current state of the art, their models and existing
models have some implications. Langevin et al. assessed the need for future work in the field as
inclusion of expanded individual-level behavior determinants, integrating interdependent
behaviors with hierarchical assessment of adaptive actions, integration of non-thermal cues to the
models (i.e. coupling of thermal cues with other environmental cues such as lighting, and etc.).
A previous study showed that an unrestricted behavior increased HVAC energy consumption since
occupants were allowed to use personal fans, heaters, thermostat adjustments, window opening
interactions [199]. There are many studies on occupant behavior, especially human-thermostat
interactions in residential buildings that reveal how humans’ interactions with thermostats vary
and modeling of these variations would result in reduction in energy consumption [13,46,145,200–
203]. Fabi et al. observed occupant interactions with thermostats in residential buildings for
deriving the physical drivers (i.e., external determinants such as indoor relative humidity, outdoor
temperature, solar radiation, time of the day) of residential heating demand. They classified the
end-users based on their interaction patterns with thermostats as active, medium and passive [13].

42
However, occupant interactions with thermostats have not been studied thoroughly in commercial
buildings.
Buildings should enable occupant interactions to choose their preferred conditions. Otherwise,
thermal discomfort increases with missing, ineffective, inappropriate, or unusable controls [155].
It is worth mentioning that the previous studies also pointed out the personalized conditioning
systems (to create a microclimate around a person) as a potential mean for increased occupant
comfort and satisfaction, and decreased energy consumption [204–206]. A climate chamber study
on personalized heating control showed that user interaction resulted in slightly less energy
consumption than fixed and automatic control [205]. Another study focused on personal heating
and confirmed lowered energy consumption and improved thermal comfort [39]. Previous studies
to understand adaptive comfort through the use of personalized conditioning systems conducted
experiments in semi-controlled environments and provided fans, shading devices, and windows
[207, 208]. They concluded that providing these options improved thermal comfort. Another study
also highlighted the improved perceived air quality through personalized ventilation [38]. Another
study provided individual air flow rate control through personalized ventilation system. Their
findings showed that 90% of the participants perceived the thermal environment as acceptable
irrespective of the ambient and personalized ventilation air temperature [37]. A previous study
focused on adaptive thermal behavior of residential settings during the summer. Providing multiple
interaction options including ceiling fans, air conditioning systems, coolers, they found almost
84% of the ceiling fans were on, and coolers and air conditioners were interacted less (1-10%)
[36]. Another study conducted in a climate chamber provided adaptive interaction options (e.g.,
clothing, desk fan, coolers, air conditioner) and concluded there exists a positive correlation
between perceived work-performance and perceived thermal satisfaction [209]. Thus, in our

43
studies, in addition to access to thermostat, we provided personal ventilation and heating as means
of potential interactions. Through this approach, we anticipate capturing the granular information
regarding occupant-heating/cooling remedies interactions.
Limitations:
➢ Previous studies looked into user behavior and behavior patterns. Some mathematical
models were developed for thermal comfort and energy quantification purposes. However,
as Santin also pointed out [46] there is a need for investigating the causes of behaviors.
Current state of art is addressing the behavior itself by investigating some patterns and user
classifications, and yet they are missing the causes of behavior (cognitive variables), which
in turn points out the necessity of how and why occupants make decisions to interact with
building systems.
➢ Not only understanding thermal behavior with causalities but also understanding thermal
behaviors through perceptions needs to be extended by coupling different environmental
drivers such as illuminance through lighting and shading systems, CO2 and noise levels
(i.e., acoustics-related drivers) [35].
➢ Most of the studies focused on human-thermostat or thermal environment control
interaction in residential buildings, there is a need for assessment of these interactions in
commercial buildings.
4.7. Human-Building Interactions in Multimodal Discomfort Context
Human-building interactions in multimodal discomfort context refers to understanding occupants’
interactions in the co-presence of visual and thermal cues (i.e., multimodal cues). Especially
understanding occupants’ interactions to multimodal discomfort requires understanding

44
occupants’ responses to visual and thermal discomfort together. Motivated by the large energy
shares of HVAC and lighting systems, in our studies we focused on system outputs of
heating/cooling remedies (i.e., temperature in our context) and lighting fixtures (i.e., light intensity
in our context) to create multimodal discomfort stimuli (i.e., solar heat gain and glare).
Occupants like having daylight in their workplace but excessive luminance could cause discomfort
and lower occupants’ productivity [210]. Glare and inadequate illuminance were identified as two
visual discomfort factors [211], and glare was depicted as ‘the harder to solve problem’. Major
source of dissatisfaction and interactions with shades is glare [212, 213]. Glare is a discomfort
determinant influencing visual comfort in daylit environments. The absence of glare is associated
with visual comfort, positive effects on health, well-being, circadian rhythms, productivity, mood,
alertness, and so on [214]. There are two major types of glare: discomfort and disability glare.
Discomfort glare takes place when the light sources do not necessarily interfere with the visual
performance of human eye but creates a discomfort feeling [215]. Disability glare is an
instantaneous physiological phenomenon that causes the reduction of visual performance, hence
visual discomfort [215]. Disability glare could be due to direct solar insolation in south and east-
facing offices [216], secondary direct solar insolation from facade reflections [216], or ambient
light reflection [188]. Occupants physically respond to glare through adjusting or closing the
shades, and response time is glare-intensity dependent [217]. In our study, we provided reflected
disability glare created by solar insolation from windows on the work plane (i.e., monitor) as a
visual discomfort factor. Using disability glare enabled us providing instantaneous initial condition
discomfort to quantify occupants' responses through their perceptual decisions since occupants
tend to avoid glare in offices.

45
Previous studies suggested that daylighting and electric lighting are inseparable [218, 219]. A
previous study emphasized that given the option to adjust artificial and daylighting, in a glare free
office, occupants dim the electric lighting and also utilize the daylighting [220]. Despite the efforts
of automating light switches, and blind controls, recently Gunay et.al. highlighted the importance
of understanding occupants’ adaptive behaviors. They emphasized the occupants’ tendency to
override automation [182], close the blinds and turn on the lights as needed (i.e., adaptive
interactions [221] in response to discomfort glare and insufficient luminance). Likewise, another
study confirmed the necessity of understanding occupant-shading interactions, satisfaction,
opinions and their consistency with the current metrics for more interactive future solutions [222].
Another study showed that on sunny days, occupants use blinds to avoid glare, and on overcast
days use blinds to get more daylight [174]. Other studies mostly focused on modelling and
simulating visual discomfort through glare estimations, not understanding occupant response to
glare through physical interactions [175, 211, 223]. Consistently, a previous study highlighted
that the significant impact of glare control on energy use intensity should motivate the future
studies to focus on glare. Especially the discomfort glare in offices, on computer work stations
have been neglected [224].
Building façade is the interface between indoors and outdoors, influential both on visual and
thermal comfort of occupants [225]. Glare and overheating due to solar heat gain could be caused
and prevented by the building façade. Incorporating the strengths of dynamic facades (e.g.,
intelligent facades, solar facades, high performance glazing) could potentially improve the
occupant comfort and decrease the energy consumption [226]. Thus, most of the studies focused
on these high-performance facades for balancing the thermal and visual environments, and indoors
and outdoors [226–230]. Yet, the most critical phase of this intelligent transformation is

46
understanding occupants’ responses to multimodal sensory cues (i.e., thermal and visual) in the
nexus of visual (i.e., glare) and thermal (i.e., solar heat gain) discomfort factors for adaptive and
resilient future designs.
Limitations:
➢ Majority of the previous studies are focused on one sensory discomfort at a time.
➢ Studies that wanted to explore thermal-visual comfort nexus tended to be limited to high
performance facades, and optimization in this regard. However, implementing high
performance façade techniques require the initial step of understanding occupant
interactions in the nexus of thermal-visual cues. Otherwise, they will have a carry-over
effect due to the oversimplified assumptions in the current state-of-art that underestimate
the occupancy impact on design by mostly focusing on only presence and schedules.
There exist not many studies to understand occupants’ thermal and visual responses together. A
previous study followed design of experiments methodology for finding the most influential
simulated occupant interactions (i.e., blinds, lights, windows, set points, fans, clothing) on thermal
sensation [7]. Despite including lighting and daylighting systems as interaction options, their
simulation provides insight on which interactions are influential on thermal sensation, but not
visual sensation. Yet, their findings showed that the energy consumption is underestimated, and
thermal comfort is overestimated when occupant behavior is not plugged into the simulations. In
our studies, we provided multimodal discomfort stimuli (i.e., solar heat gain and glare) to
understand the occupants’ responses with a holistic approach (i.e., providing the thermal and visual
realism and means of interactions).

47
4.8. Human-Building Interactions In Multi-Occupancy Context
Presence research aims to understand what leads to people’s sense of ‘being there’ in virtual
environments when they are aware of the environment’s being synthetic [231, 232]. Social
presence is a computer mediated communication design principle and defined as the feeling of
being with someone else in a computer aided environment [233]. Copresence differs from social
presence; copresence represents a psychological bond while social presence implies the perception
of the medium’s ability [234]. Copresence is two-folded; mode of being with others and sense of
being with others [233]. Copresence studies tried to understand how to enhance the feeling of
being with someone else in a virtual environment, and also interested in involving some form of
social interactions (e.g., collaboration) [235, 236]. A previous study stated that for social presence,
humans need access to virtual party’s intelligence [231]. Likewise, another study argued that
perception of social entity takes place with some level of interactivity and plausible responses
[237]. A study on responses of people to virtual humans in immersive virtual environments showed
that although the participants knew that the agents were computer generated, the participants with
higher social anxiety were more likely to avoid disturbing them. They indicated that this suggest
that people perceive the avatars as social actors even in the absence of mutual interactions [238].
It is worth mentioning that their study took place in an unstructured context (i.e., participants were
free to move around, explore the virtual environment and interact with the virtual humans).
So far, we have covered the current state of art with regards to human-building interactions at
interaction-type level. Most of these studies focused on interactions in single occupancy settings.
However, a through contextual understanding of occupant interactions requires exploring the
interactions in multi-occupancy spaces as well. Otherwise, occupant behavior models and related
energy simulations will be oversimplified.

48
Previous studies on social behavior in occupancy context are limited to social behavior modeling
of pedestrians, and emergency evacuation [239]. In these regards, macroscopic models focusing
on crowds through pedestrian density and velocity, and microscopic models focusing on individual
pedestrians (e.g., social force model [240], cellular automata [241]) were developed. Yet,
microscopic understanding of occupant behavior in multi-occupancy offices context is almost
untouched, and there are only few studies incorporating multi-occupancy. Previously, Haldi and
Robinson found that occupants in single occupancy offices tend to interact with the blinds more
frequently and have lower light intensity compared to multi-occupancy cases [242]. They
performed a field study in fourteen south-facing offices; six had two occupants in which one
occupant had access to blinds, eight were single occupancy offices. This study was limited to
understand blind interactions while the occupants had horizontal workspace. However, the results
are encouraging to unveil the unknowns of occupant interactions in contextually-rich multi-
occupancy settings by providing both heating/cooling remedies and different lighting fixtures.
Another study [243] performed occupant interviews regarding their interactions with controls in
high performance buildings. This study reported that social concerns (i.e., occupant’s inactivity
due to hesitance to affect others) is a mentioned reason (the study reported only one occupant’s
answer with regards to social concerns) for not interacting with blinds.
A previous study highlighted the potential influence of multi-occupancy on patterns of each
occupant’s interactions [14]. Addressing the multi-occupancy interactions is not simply deriving
the interaction occurrence probabilities for both occupants [14]. It is a deeper process requires
understanding the psychological and social influences of the presence of other occupants on
interactions [100]. Thus, in this dissertation, we both focus on metrics attributed to occupant
interactions (i.e., type, kind, number, response time and patterns of interaction decisions) and

49
presence of other occupants (i.e., co-presence and social avoidance), as well as the individual sense
of presence.
Limitations:
➢ Due to the complexities and dynamicity of occupant behavior, previous studies prioritized
and were limited to understanding occupancy in single occupant offices. Not even occupant
behavior in single occupancy offices is thoroughly understood due to oversimplified
assumptions on behavior. However, exploring the occupant behavior in multi-occupancy
and single offices simultaneously is possible through quantifying the metrics attributed to
behavior (e.g., number, type of interactions).
➢ Existing models frequently focus on single occupancy workplaces, rarely consider multi-
occupancy [14]. It is imperative to potentially address the presence of another occupant in
behavior models. The first step for it will be retrieving the information regarding occupant
behavior in multi-occupancy context.
Currently, quantification of multi-occupancy interactions is an unaddressed problem with a
potential rewarding contextually-rich occupancy modeling [14]. To address this problem, we
explored the influence of the presence of an office-mate on occupants’ interactions in virtual
environments. In this dissertation, we also explore the human-building interactions in multi-
occupancy settings. For this, we immersed the participants in structured contexts (i.e., controlled
experiments) wherein they had a virtual (i.e., avatar) office-mate with realistic interactive features
(e.g., working on his laptop and the participant could clearly see his typing movements), but no
responsiveness since we were interested in the influence of psychological bonding (i.e., co-
presence) on human-building interactions.

50
4.9. Energy and Comfort Implications of Human-Building System Interactions
Occupants respond to environmental conditions when to retain or maintain their comfort [244–
246]. Understanding how human operators (i.e., occupants) interact with building systems and
why they decide to interact with these systems are important for developing behavior models to be
used in simulations for increased comfort and energy savings, as well as quantifying the energy
implications of these interactions [12, 103]. In the previous section, current state of art on human-
building system interactions was surveyed and this section covers the energy and comfort
implications of these interactions to highlight the importance of understanding how and why
occupants perform energy consuming HBIs.
4.9.1. Energy and comfort implications of human-blinds and lighting systems
interactions
Occupants’ responses to lighting and daylighting systems influence their comfort and affect the
building energy consumption [221, 247]. Occupants’ adaptive behaviors or actions are mostly
observed when they want to adjust the environment to their desired comfort levels, which usually
takes place in discomfort conditions [248]. Occupants’ visual discomfort may result in their
interaction with shading and lighting systems [179]. Although glare and insufficient illuminance
have been depicted as two fundamental causes of discomfort, due to the involvement of multiple
variables, determinants of visual discomfort have not been thoroughly discovered [211].
Commercial building lighting systems are usually operated based on pre-set occupancy schedules
and lighting level standards, which do not include in the real needs and behavior of occupants
[249].
Incorporating algorithms for occupant behavior, Newsham showed that occupants’ comfort
increased when they were provided manually controlled shading devices compared to the absence

51
of shading devices. Although when manually controlled lighting systems were included, increased
blind use resulted in increased energy consumption, total energy consumption was lower than the
environments in which lighting level was kept constant (i.e., no occupant control over lighting
levels). With the same study Newsham also concluded that, understanding the influence of
occupant behavior through manual operation of both lighting and blind systems require spatial
definition of the occupants (i.e. where occupants are located in the environment) [249]. Mahdavi
et al. showed that 66-71% lighting systems-related electricity energy savings is possible through
implementation of occupancy sensors and daylight-responsive dimming devices [179]. Gunay et
al. showed that adaptive occupant learning blind and lighting systems would result in decreased
energy consumption and approximately 80% reduction in occupant overriding to automated
systems [184]. It is worth mentioning that blinds could be accounted as building’s both thermal
and daylighting systems. Inappropriate use of blinds could increase electricity consumption and
heating/cooling loads in a building [250]. Thus, in our studies we focused on multimodal stimuli
(i.e., thermal and visual through solar heat gain and glare) to account for the interrelation between
thermal and lighting/daylighting systems interactions.
4.9.2. Energy and comfort implications of human- thermostat interactions
One of the factors influencing occupants’ thermal comfort is thermal sensation. In this dissertation,
we investigated the relation between thermal perception and related adaptive actions through
perceptual decision-making processes. Previous studies showed that thermal comfort-related
internal and external determinants vary over time [251, 252] and individual differences (e.g. age,
gender, percentage of body fat, metabolism, clothing) contribute to the complexity of developing
generalized thermal comfort models [11, 167–169, 253, 254]. Although Fanger’s thermal comfort
model is still being used, it does not address the personal thermal preferences of humans [11].

52
Existing models’ inability to capture and integrate individual differences resulted in inaccurate
estimation of occupant comfort and energy consumption and led to a more human-centric
modelling approaches (e.g., Human Centered thermal Comfort Identification (HCCI) through
which direct measurements of perceptions and physiological changes from humans were enabled
[171], and human-in-the-loop approach that saved 30% in heating load and 38% in cooling load).
Moreover, recent studies have shown that for addressing the differences, comfort could be
addressed through human perception and sensation via local sensing (e.g., via skin temperature)
[255]. Zeiler et al. proposed a human-in-the-loop approach for occupant control over HVAC
systems through which they addressed physiological variations across body parts for identifying
personal perceived comfort. Their study showed that there is a significant potential to overcome
comfort and energy implications by their methodology [11].
For addressing the energy and comfort implications of HVAC controls, Mo and Mahdavi proposed
an agent-based simulation approach [256]. Through different HVAC control strategies, Matthews
et al. showed that saving 30% of total building energy is possible while improving occupant
comfort [257]. Van Oeffelen et al. showed that by optimizing human comfort differences between
individuals, it is possible to save 25% of energy [258]. Ari et al. showed that through occupants’
thermostat interaction patterns, it is possible to derive occupants’ preferred temperature and
temperature tolerances which resulted in improved thermal comfort and energy savings [259]. In
addition to these, Moerstra and Beuker deduced that occupants of office buildings feel more
comfortable when they have more control over heating-cooling systems [13, 153, 155].
In addition to separate analyses of lighting/daylighting and HVAC systems, there exist some
studies covering the energy and comfort impacts of coupled lighting-daylighting and thermal
building systems. A previous study showed that in cooling-dominant climates, through energy

53
efficient window design, it is possible to reduce solar heat gain while keeping the occupants
satisfied with their indoor environments [229]. Newsham also found that implementing manual
control availability to both blinds and artificial lighting systems reduced solar heat gain which
resulted in decreased cooling energy (by 7%), increased heating energy (by 17%), and increased
lighting energy (by 33%) due to the decrease in daylight [249]. Another study also observed
cooling load reduction of 7-32% and substantial amount of daily energy consumption reduction
through use of dynamic daylight controls [191]. Moreover, Roetzel et al. mentioned that a holistic
optimization study is necessary for leveraging lighting/daylighting and HVAC building systems
for better comfort and better building energy performance [260]. Another study incorporated
occupant behavior in simulations; analyzed lighting, plug-loads, HVAC, windows impact
separately and together, and achieved up to 22.9% and 41% energy savings respectively [261].
Limitations of energy and comfort implications of human blinds, lighting and thermostat
interactions:
➢ Several studies highlighted the significant gap between the predicted energy performance
of buildings and their measured actual performance. Uncertainties regarding behavior of
building occupants are one of the key factors limiting the ability of energy simulation tools
to accurately predict real building energy requirements.
➢ The above-mentioned studies show that there is a need for real occupancy data for
assessment of environmental, personal and contextual factors influencing comfort and
energy consumption. O’Brien also emphasized this need for empirical studies through
which a database could be created [100].
While discussing the energy saving potentials, studies often focused on one or two energy-related
behavior [262]. Coupling of human-system interactions and providing multiple interaction means

54
(e.g., electric lighting and blinds) to occupants are needed for hierarchical assessment of
occupants’ adaptive behaviors with causalities reflecting how and why occupants interact with
their indoor environments. Despite the emphasis on the determinants of human-systems
interactions and determinants of energy consumption behavior, they have not been integrated into
decision-making models for understanding human behavior. Even partial consideration of these
determinants in surveyed studies showed substantial energy savings and comfort improvement,
which shows that development of high-resolution occupant behavior models incorporating these
determinants and coupled human-system interactions through their response to the perceived
changes in environmental cues has potential to save energy in buildings. Thus, it is imperative to
first understand how and why occupants interact with their indoor environments.

55
Chapter 5. Objectives and Research Questions
Objective 1: To test the adequacy of using virtual environments in the context of thermoception
for human-building interaction studies.
Research question 1.1: How to understand the adequacy of using immersive virtual
environments with regards to thermoception using human-building interaction attributed
factors (e.g., number and type of interactions) as benchmarking metrics?

Objective 2: To investigate human-building interactions (specifically with heating/cooling
remedies and lighting/daylighting systems) by understanding occupants’ perceptual decision-
making processes under multimodal environmental stimuli (i.e., thermal and visual stimuli) in
single and multi-occupancy offices.
Research question 2.1: What are the occupants’ responses (i.e., interactions with
heating/cooling remedies and lighting fixtures) to multimodal stimuli through their
perceptual decisions in single and multi-occupancy offices?
Research question 2.2: What are the reasons of occupants’ perceptual decisions to interact
with heating/cooling remedies and lighting fixtures under multimodal discomfort in single
and multi-occupancy offices?

56
Chapter 6. Benchmarking Thermoception in Virtual Environments to Physical
Environments for Understanding Human-Building Interactions
Occupant-building interactions are complex, multi-layered and influence both the environment
and occupants [263]. These interactions, defined as any response of an individual or groups to their
environment [5], are hard to predict due to the stochastic nature of human behavior [103]. Yet,
occupants’ interactions with building systems and elements and occupant related factors (e.g.,
behaviors, preferences) substantially influence satisfaction and comfort, as well as a building’s
energy consumption [4, 248]. One of the most comfort- and energy- influencing interaction type
takes place in the context of thermal comfort (e.g., interaction with heating, cooling, ventilation
devices). Thermal comfort is defined as the state of mind that reflects satisfaction with a thermal
environment and is usually assessed through subjective evaluations, such as thermal vote
assessments in physical environments [264]. There are many contextual factors that affect
occupant behavior [100, 248]. Several decisions made during the design stage of a building (e.g.,
control options, interior design, space orientation), influence occupants’ thermoception (i.e., the
sense by which animals perceive the temperature of the environment and their body). Due to the
limited resources for testing the influence of such contextual variables on comfort and satisfaction,
it might not always be feasible to use physical mock-ups during the design stage of a building in
order to assess the impact of decisions on occupant’s thermoception. However, virtual
environments (environments created with virtual reality technology) could provide unique
opportunities to construct such variables easily and relatively inexpensively. This could facilitate
various occupant behavior studies in built environments for unbuilt spaces or for existing spaces
that go through renovations. Obtaining the micro-level information regarding occupant behavior
could be used in user-centered designs or user-centered building operations [5].

57
In order to collect thermal behavior data in built environments, there exist two contemporary
methods [265]: creating acclimatized climate chambers, which provide experimental control and
increased internal validity [266] and using real buildings as test beds, such as in [265, 267] and
[268]. The former method might not provide the mundane realism due to the lack of contextual
factors, which do exist in real settings [269]. In addition, not many design teams have access to
climate chambers to test their design decisions. Data collection in real test beds is advantageous
for providing the realism with ‘real occupants’ and ‘real office environments’ with contextual
variables and for long-term studies to understand temporal factors’ influences on the
thermoregulation behavior. However, the drawback is physical mock-ups have to be built to test
different design decisions during the design stage, which is prohibitively expensive and, in many
cases, might not be feasible. As de Dear et al mentioned, in the past 20 years, neither method has
been used as frequently as comfort simulations due to the growing availability of simulation tools.
Yet, these simulation studies mostly lack the ground truth (i.e., lack of human subject data) [265].
Virtual environments provide an egocentric multimodal sensory (i.e., visual, haptic, auditory,
olfactory, thermal, gustatory) [40] experience to humans wherein visual, auditory and kinesthetic
aspects are defined by computers [270]. Humans are immersed in virtual environments wherein
their perceptual systems interact with a simulated synthetic information through displays. This
synthetic information is conveyed to the users through their perceptions as if it were real; this
information envelopes them perceptually while continuous streams of stimuli are present [106].
Virtual environments have been widely used in many disciplines where human experience is in
the foreground (e.g., social psychology [40], medicine [271], education [272] and training [273],
design [274, 275] and engineering [276]. These environments provide experimental supremacy in
mundane reality in-situ controlled experiments to isolate the exogenous factors and stimuli to

58
transfer behaviors observed in virtual environments to the physical environments. Thus, virtual
environments are alternative venues for human behavioral studies as controlled scenarios could be
created and tested relatively easily [40] and in many cases cost effectively compared to physical
mock-ups, especially in the case of built environments.
There exist several human-centered experimental studies in physical environments for
understanding and improving occupant thermal comfort (e.g., [165, 277–279]). However,
integrating thermal cues (through thermoception) to virtual environments for understanding
occupant behavior and improving perceptual realism in virtual environments have not been well
studied yet. Realistic virtual built environments could potentially enable us to test different design
decisions, such as the impact of changing the orientation of an office and the interior design or the
material properties of objects (walls, furniture, etc.) on occupants’ thermal comfort. Thus, the
objective of this study is to benchmark virtual environments to physical environments with regards
to thermal stimuli, by comparing users’ perceived thermal comfort and satisfaction. In this study
we conducted a human subject experiment in an acclimatized hybrid environment where we
recorded participants’ (physical, psychological) responses when interacting with heating/cooling
remedies. In order to understand the adequacy of using virtual environments in the thermoception
and human-building interactions domains, we compared participants’ thermal comfort and
satisfaction between virtual and physical offices.
6.1. Methodology
In order to test the adequacy of using virtual environments in the context of thermoception for
human-building interactions studies, we tested the alternative hypotheses that thermoception is
different in virtual and physical environments, yet, we expected it to be similar in the two
environments (expected to fail to reject the null hypothesis). To explore our research objective, we

59
devised a human subject experiment. Prior to the execution of the experiment, we ran a pilot study
to screen and overcome any erroneous parts (e.g., models, building systems/elements, surveys,
procedures), and make the experiment more robust to achieve the objectives of the present study.
Based on the results of our pilot study with 19 participants, model modifications were done, the
number of surveys were reduced, some of the survey questions were modified, new survey
questions were added, and minor modifications were performed on the procedure. Below is a
detailed explanation of our experimental study.
6.1.1. Design of experiment
The experiment setting was a multi-occupancy office space of 42 square meters with seven
workstations (Figure 1). Only one of the seven workstations was used for the experiment. The
virtual office space was modeled following the actual office’s properties (e.g., number and location
of workstations, office area, etc.). Each participant was recruited to two sessions: physical session
and virtual environment session. Participants were welcomed to either hot or cold indoor
conditions (randomly assigned). The physical office space was acclimatized during both virtual
and physical experiment sessions. Participants were asked to complete a generic office task (i.e.,
a reading comprehension task that was leveled for elementary school) to stay focused for
simulating the office space realism during the experiment sessions. No task performance
assessment was conducted.

60

Figure 1. Plan of the office space and location of the experiment workstation
During each session occupant-building system interactions were recorded, participants completed
surveys before and after the experiment sessions, and subjective thermal comfort and satisfaction
votes per each interaction were collected. Thermal stimuli in the virtual office were provided
through the heating/cooling remedies in the physical environment upon participant’s interaction
with them. For example, if the participant turned on the radiant heater in the virtual office, the
experimenter turned it on in the physical office at the same time. The participant did not see the
experimenter’s actions as he/she would be wearing a head mounted display for the virtual office
session. During the information session, the participants were informed that they would be able to
adjust their environment, resulting in possibly perceiving a change in the thermal condition of the
room based on their interactions.
6.1.2. Environment and apparatus
The workstation included a computer, a keyboard, a mouse, and a monitor in the physical office
space (Figure 3). A desk fan, local heater/cooler fan, radiant heater, cup of hot or cold beverage

61
(depending on the environment: hot vs. cold) were added to the workstation in addition to the
existing thermostat (i.e., to facilitate the control over the HVAC system). This modified
workstation enabled the participants to be equidistant from the heating/cooling remedies in order
to eliminate the ‘ease of control’ bias from their interactions. A virtual abstraction (Figure 2. c &
d) of the physical office (Figure 2. a & b) was also created in a similar fashion. This virtual office
space was modeled and rendered in Revit 2015, 3 DS Max, and Unity 3D. Modeling was
performed to the scale of the physical systems. Necessary interactions in the virtual environment
(e.g., adjusting the speed level of the desk fan) were coded by using C# in Unity. An Oculus Rift
Head Mounted Display was used for immersing the participants in the virtual environment. Indoor
air temperature was identified as the independent variable of the study, being one of the most
influential factors of thermal comfort. In order to isolate the thermal perceptions of participants,
season, indoor lighting configuration and intensity, context, interior design, location and
orientation of the workstation, ambient factors (except for the indoor air temperature), and
participants’ activity levels were kept constant at all times. During the experiment, indoor air
temperature was recorded by SMAKN DHT22 AM2302 digital temperature and humidity
measurement sensor (with temperature accuracy of +-0.5C and humidity accuracy of +-2% RH
(Max +-5%RH)).

62

(a) (b)

(c) (d)
Figure 2. (a) and (b) physical office space; (c) and (d) virtual office space
6.2. Experiment Details
6.2.1. Recruitment
Prior to the recruitment, the Institutional Review Board (IRB) approved the study. A total of 56
participants (34 females, and 22 males), which provided us with 99% power, 0.7 effect size, and
95% confidence, were recruited in the study. The participants were recruited through the
University’s human subject pool, whose efforts were compensated through course credits. If any
participant experienced motion sickness anytime during the experiment, their credits were granted,
but they were dismissed from the study and their data was not included. Motion sickness was

63
determined either through self-reporting during the experiment session, or through the simulator
sickness surveys.
6.2.2. Pre-experiment session
Information session took place in a room (different than the experiment office) with normal indoor
air temperature. During the information session, each participant was required to read the IRB-
approved consent form with the necessary experiment details, procedure, risks and rights listed. In
this room, the participants also completed a pre-experiment survey regarding their demographics
and other personal information. Participants were required to complete the simulator sickness
questionnaire prior to starting the virtual session. The pre-experiment session took 15-20 minutes.
We used the pre-experiment session as a way for the participants to acclimatize to the normal
indoor air temperature and neutralize any effect from their prior comfort state. Afterwards, each
participant was introduced to the experiment setting.
6.2.3. Experiment session
The experiment was conducted in two groups based on indoor acclimatization: (1) cold condition
(18-Celsius degrees) and (2) hot condition (28-Celsius degrees) in physical and virtual offices.
These perceivable non-extreme cold and hot temperatures were chosen as adequate representations
of cold and hot thermal conditions and are sufficient to provide discomfort stimuli without
extremity. It is worth mentioning that these cold and hot conditions were set as the initial
conditions; indoor air temperatures changed as the participants interacted with the heating/cooling
remedies. In other words, the dynamicity, heterogeneity, and environmental realism of indoor
environments were accommodated through the integration of participants’ interactions with the
heating/cooling remedies. For example, participants could turn on the heating, ventilation, air
conditioning (HVAC) system, turn it off, adjust the set point, adjust the radiant heater and so on.

64
If these interactions had air temperature or air velocity outcomes, the participants experienced this
environmental dynamicity as a natural outcome of their interactions. This way, the contextually
rich virtual environment provided participants with interaction means and controls. Participants
could see the thermostat, fan, heater, cooler, glass of beverage, in the virtual office and interact
with these means through the controller. If there was an animated change (e.g., the oscillating desk
fan), they could also see this dynamic visual change, and if there was a thermal outcome, they
could also perceive it. In order to understand complex occupant behavior in human-building
interactions in the virtual built environments context, it is imperative to provide enhanced
perceptual realism. In other words, if the participant interacted with the oscillating fan, saw the
oscillation but did not perceive the ventilation, we would not be able to assess this interaction as
an adaptive response due to lacking realism. In this context, the notion of interactions follows the
psychological explanation of the activity theory (equipment mediates the activity connecting the
individual to the environment) [280]. Tools shape experience and the activity (interactions in this
study) cannot be interpreted without understanding the role of artifacts (i.e., heating/cooling
remedies) in everyday use [281].
The order of physical office and virtual office sessions were randomized for each participant. Each
day, participants were recruited in either hot or cold starting conditions for almost an hour in order
to eliminate the bias related to the outdoor air temperature. In addition, temporal factors were also
eliminated through expansion of the recruitment schedule to a wide band ranging from 8 am until
11 pm. During the experiment, participants were provided with equal number of interaction options
both in the virtual and physical offices: (1) adjusting the local heater/cooler, (2) drinking hot/cold
beverages, (3) adjusting the desk fan, (4) adjusting the thermostat, (5) adjusting the radiant heater
(Figure 3), and (6) adjusting the clothing level. In the virtual office, participants chose interactions

65
by using the controller. The experimenter simultaneously implemented the chosen interaction in
the physical office so that the outcomes would be perceived by the participant in the virtual office.
For example, if the participant were to increase the desk fan speed in the virtual office by pressing
a button on the controller, the experimenter would toggle the fan speed in the physical environment
to match their interaction. Similar to the physical office, there existed no visual change in the
virtual office when the participants interacted with the local heater/cooler, radiant heater and the
thermostat but they perceived the change in temperature as the result of their interactions (turning
on/off devices, etc.). Similarly, the fan in the virtual office was animated for all four levels of
ventilation like the one in the physical office. The clothing level of the participants was not leveled.
They were informed to start both virtual and physical office sessions with the same initial clothing
level (per participant). Adjusting the clothing level option was provided because a previous study
showed that providing flexible clothing level instead of fixing it decreased the discomfort [282].
If they had an extra layer of clothing, they could put it on, or vice versa. Otherwise, they were not
provided an extra layer.

Figure 3. Type of interactions provided in virtual (left) and physical (right) environments [(1) local
heater/cooler, (2) hot/cold beverages, (3) desk fan, (4) ad thermostat, (5) radiant heater]
While performing the reading comprehension tasks (i.e., 2 tasks per environment), the participants
were able to make themselves comfortable through interactions (i.e., adaptive responses). Each

66
experiment session was conducted for 15 minutes (total of 30 minutes). Participants were allowed
to interact with the building systems and heating/cooling remedies as much as they wanted. They
reported their perceived thermal comfort and satisfaction per interaction (i.e., prior to and
subsequent to each interaction) through subjective thermal comfort and satisfaction votes after
each experiment session. In addition to the perceived thermal comfort and satisfaction, perceived
indoor air temperature data were also collected using survey questions. For a systematic
comparison, six parameters were measured in this study: (1) type of occupant-system interactions,
(2) number of occupant-system interactions, (3) perceived thermal comfort prior and subsequent
to each interaction (e.g., participants were asked to report their thermal states (i.e., much too warm,
uncomfortably warm, comfortable, comfortably cool, uncomfortably cool, much too cool), (4)
perceived thermal satisfaction (i.e., extremely satisfied, moderately satisfied, slightly satisfied,
neither satisfied nor dissatisfied, slightly dissatisfied, moderately dissatisfied, extremely
dissatisfied), (5) perceived indoor air temperature and (6) immersion experience of the
participants.
6.3. Post-Experiment Session
It is worth reminding that the order of virtual and physical office sessions was randomized for each
participant. Thus, depending on which session the participant started the experiment from, the
order of post-experiment surveys also varied. In other words, there were two separate post-
experiment sessions (i.e., one after the virtual office session and another one after the physical
office session). For the post-experiment sessions, the participants were welcomed back to the
environment with normal office temperature, where the pre-experiment session took place. After
both the physical and virtual environment sessions, participants were required to complete a
thermal experience post-experiment survey. This survey was designed to retrieve information

67
regarding the perceived thermal comfort and satisfaction, type, order, causes of the interactions
and perceived indoor air temperature. Subsequent to the virtual office session, participants were
required to complete the simulator sickness questionnaire. Based on this questionnaire, data of the
participants with motion sickness was excluded from the study. They also completed a post-virtual
environment survey including the presence questionnaires (i.e., Slater-Usoh-Steed [283], and
Witmer and Singer [106]) to measure the participants’ presence in virtual office, and Immersive
Tendencies Questionnaire (ITQ) [106] to measure the presence tendencies of the participants in
the virtual office. In total, there were 54 questions, and they were all based on 7-Likert Scale. After
filling out the post-experiment surveys, the participants were thanked and dismissed.
6.4. Results and Discussion
The design of the experiment had a mixed nature of different factors (i.e., environment,
temperature) with multiple levels (i.e., physical and virtual, hot and cold). The post hoc tests on
the data showed that the majority of the sample data were not normally distributed. Thus, non-
parametric statistical analysis methods, Wilcoxon Signed Ranks tests for matched/paired (i.e.,
dependent) samples and Mann Whitney tests for independent samples were used for the hypotheses
testing. If significance values in these tests are below 0.05, there exists statistical significance
between two samples. Since the participants were randomly assigned to hot and cold
microclimates, in order to make sure there are no group differences, we compared the age of
participants in hot to that of cold, and likewise gender in both conditions using Mann Whitney
Tests. There existed no statistical difference between age (p=0.9>0.05) and gender (p=0.278>0.05)
of the participants between hot and cold conditions. We identified multiple metrics (i.e., perceived
comfort, satisfaction, temperature, number and type of interactions) that could be used for

68
comparing thermoception in physical and virtual environments. Below is a summary of the
findings.
6.4.1. Actual versus perceived indoor air temperature
We analyzed the indoor air temperature sensor data both for the physical and virtual offices to
make sure they were not statistically different at the beginning of the experiments. The initial cold
condition in the physical office was not significantly different than the initial cold condition in the
virtual office (p=0.073>0.05) and likewise in the hot condition. Thus, the environments were not
significantly different (p=0.925>0.05). Then, we compared the actual indoor air temperature to the
perceived temperature. If the participants were able to correctly identify the indoor air temperature
within the range of 3 Celsius degrees, they scored ‘1’, else ‘0’. Based on this binary categorization
of participants’ ability to quantify perceived indoor air temperature (which is the product of their
mind based on thermal cues), the Wilcoxon Signed Ranks test was performed on the data collected
in the physical and virtual offices for hot and cold conditions. Significance values for these tests
were found to be larger than 0.05 prior to the interactions in the cold condition, hot condition, and
subsequent to the interactions in the cold and hot conditions (i.e., 0.366, 0.527, 0.405, 1,
respectively). Hence, no statistical significance was observed. Thus, we failed to reject the
hypothesis that “people’s ability to quantify the perceived indoor air temperature is different in the
physical and virtual environments.” Participants’ ability to quantify the perceived indoor air
temperature was poor in both the virtual and physical environments, which were not found to differ
from each other.

69

Table 2. Percentage of correctly and incorrectly (lower or higher than it is) reported perceived
temperature
This inference does not mean that people perceive temperature the same in the virtual and physical
environments. In order to clarify this point, it is worth comparing the percentage of participants,
who correctly and incorrectly quantified the temperature (i.e., perceived temperature higher or
lower than it is) in the physical and virtual offices in hot and cold microclimates (Table 2). The
percent values show that in both offices (virtual and physical) and under both conditions (hot and
cold), the majority of the participants were not able to quantify the indoor temperatures correctly.
The percentage of participants who could correctly quantify the actual temperature is slightly
higher in the virtual office than the physical office under both hot (42.86% in virtual vs. 35.71%
in physical) and cold (39.29% in virtual vs. 28.71% in physical) conditions prior to the interactions.
However, this was not the case for subsequent to the interactions: more participants in the physical
office under cold condition (42.86%) quantified actual temperature correctly than the virtual office
in cold condition (32.14%). The percentages for correct quantification of indoor air temperature
subsequent to the interactions under hot condition are the same for physical and virtual offices
(10.71% each).
COLD CONDITION HOT CONDITION
Correct Lower
than it is
Higher
than it is
Correct Lower
than it is
Higher
than it is
Prior to
interactions
Physical 28.71 42.86 28.57 35.71 42.86 21.43
Virtual 39.29 35.71 25 42.86 39.29 17.86
Subsequent
to
interactions
Physical 42.86 10.71 46.43 10.71 89.29 0
Virtual 32.14 7.14 60.71 10.71 82.14 7.14

70
However, an interesting observation is, prior to the interactions, both under the cold and hot
conditions in both virtual and physical offices, the percentage of participants feeling the
temperature ‘lower than it is’ is higher than ‘higher than it is.’ This shows, prior to their
interactions, people tend to perceive temperature lower than it is, regardless of the environment
being virtual or physical. However, this is not the case subsequent to the interactions. In the cold
condition, 46.43% and 60.71% of the participants in the physical and virtual offices, respectively,
thought the indoor air temperature is higher than it actually was, subsequent to their interactions.
These results show that after the participants modified the environment to make them warmer in
the cold condition, they perceived the indoor air temperature higher than it actually was. Similarly,
in the hot condition, 89.29% and 82.14% of the participants in the physical and virtual office,
respectively, perceived the indoor air temperature lower than it actually was subsequent to their
interactions. There is an agreement; people perceive the indoor air temperature as lower than it is
(if the initial condition was hot) and higher than it is (if the initial condition was cold) subsequent
to the interactions. This is potentially due to the perceived relief and improved perceived thermal
comfort of participants, followed by their physical adaptive responses.
6.4.2. Thermal comfort and satisfaction
An agreement in the subjective votes in both offices indicates the similarity of physical and virtual
environments with regards to the perceived thermal comfort and satisfaction. In order to compare
the subjective reporting of comfort and satisfaction, we defined two delta terms (for comfort and
satisfaction), which represent the direction of overall comfort and satisfaction experience. Delta
terms were computed based on the initial and final subjective votes of the participants by finding
the difference between the subjective thermal votes subsequent to the last interaction (i.e.,
representative of the final perceived thermal comfort and satisfaction) and prior to the first

71
interaction (i.e., representative of the initial thermal comfort and satisfaction that resulted in
adaptive interactions). The delta comfort in the physical and virtual offices for the hot and cold
conditions were compared by the Wilcoxon Signed Ranks test. The results showed that there is no
significant difference between the physical and virtual offices with regards to the perceived
comfort (in hot condition p=0.739>0.05, in cold condition p=0.705>0.05) and satisfaction (in hot
condition p=0.180>0.05, in cold condition p=0.826>0.05). Thus, we failed to reject the null
hypotheses, as the perceived thermal comfort/satisfaction in the virtual environment was not found
to differ from perceived thermal comfort/satisfaction in the physical environment.
In addition to the delta comparisons, thermal comfort and satisfaction prior and subsequent to the
interactions in the physical and virtual offices were compared. The results show that the comfort
prior to the interactions in the hot physical and virtual offices is not significantly different
(p=0.248>0.05). Subsequent to interactions is also not significantly different (p=0.317>0.05). The
same is observed in the cold physical and virtual offices: comfort prior to the interactions
(p=0.854>0.05) and subsequent to the interactions (p=0.705>0.05) is not significantly different.
Satisfaction prior (p=0.755>0.05) and subsequent to (p=0.755>0.05) interactions in cold condition,
and prior (p=0.357>0.05) and subsequent to (p=0.755>0.05) interactions in hot condition in
physical and virtual offices is also not significantly different. The hypotheses tests, on comfort and
satisfaction prior and subsequent to the interactions in both environments under hot and cold
conditions separately, show that delta terms were adequate metrics for the comparison. Moreover,
these results also show that there exists no difference in comfort and satisfaction prior or
subsequent to the interactions in physical and virtual offices. Thus, the subjective thermal comfort
and satisfaction assessment methods (i.e., subjective votes) could be used as markers of
thermoception in virtual environments.

72
6.4.3. Number and type of interactions
In order to understand the adequacy of using virtual environments in the thermoception domain,
we also measured quantifiable objective metrics (i.e., number and type of interactions) related to
the participants’ thermoception. We hypothesized that the number and type of interactions in the
virtual and physical environments are similar. In the post-experiment surveys, the participants
were asked to report the type and order of interactions they performed, as well as the reasons that
caused them to perform each interaction. ‘Corrected number of interactions’ were used for
analysis. From the total number of interactions, we subtracted any interaction that is not an
adaptive physical response for adjusting the thermal conditions, such as interactions performed
only to explore the virtual environment. The analysis below shows our results.
Type of Environment COLD
CONDITION
HOT
CONDITION
Mean Standard
deviation
Mean Standard
deviation
Physical 3.57 2.27 3.39 2.71
Virtual 2.43 1.95 3.36 1.79

Table 3. Mean and standard deviation for the number of interactions under both hot and cold conditions
in virtual and physical offices
The mean number of interactions under both hot and cold conditions in the virtual office are lower
than that of in the physical office (Table 3). A Wilcoxon Signed Rank test was performed to
compare the number of interactions in the virtual office to physical office. The results showed that
the number of interactions in the virtual office is not significantly different than that of in physical
office (p=0.051>0.05). Thus, we failed to reject the hypothesis and concluded that the number of
interactions in the virtual environment is unlikely to differ from the physical environment.
However, the significance value is marginal. Then, the hypotheses tests were performed on the

73
number of interactions in hot and cold conditions, separately. The results show that the number of
interactions in the virtual office is significantly lower than in the physical office under cold
conditions (p=0.013<0.05). Under hot conditions, the number of interactions in the virtual office
is not significantly different than in the physical office (p=0.963>0.05). Despite the similarity
under the hot condition, the mean number of interactions in the virtual office is still lower than in
the physical office. A potential reason for fewer interactions in the virtual office could be that
participants might not have thought their interactions in the virtual office would affect the indoor
air temperature. However, this does not explain the significant difference in the cold conditions.
Another potential explanation could be the presence or lack thereof. In order to assess the
relationships between presence and its potential markers (i.e., comfort, satisfaction, number/type
of interactions, perceived temperature), consecutive Pearson product-moment correlations (i.e., r-
values) were computed. There is a strong, positive correlation between mean Witmer-Singer
presence and comfort prior to the first interactions in the hot virtual office (i.e., r=0.704, p<0.0001),
which shows that increases in perceived thermal comfort prior to interactions in the hot virtual
office is correlated with increases in reported presence. There exists significant positive correlation
between this comfort variable and the three factors of Witmer-Singer presence (involvement
(r=0.58, p=0.001), sensory fidelity (r=0.686, p<0.0001), immersion/adaptation (r=0.591, p=0.001)
in the hot virtual office. Such relations do not exist in the cold virtual office. Not observing such
correlations in the cold virtual office could be a potential presence-related reason for the significant
difference in the number of adaptive interactions in the cold virtual office.
In addition to the number of interactions, virtual and physical offices were also compared with
respect to the type of initial adaptive interactions under hot or cold conditions. The initial
interaction is representative of the emerging response for one’s adaptive behavior, which is the

74
immediate solution that they could think of or need. The results show that type of first interaction
in the hot virtual office is not significantly different from that in hot physical office
(p=0.659>0.05), and likewise no significant difference was detected between cold virtual and
physical offices (p=0.868>0.05). The percent distributions of type of first interaction per
environment are visualized for hot and cold conditions (Figure 4). The initial interactions’ in
virtual and physical environments being similar could be used as another marker to benchmark
virtual environments to the physical environments in the thermoception domain.

Figure 4. Percent distributions of type of first interactions in hot/cold conditions
In both office environments under both conditions, adjusting the local heating/cooling remedy has
a high percentage as an emerging adaptive interaction. In the cold virtual office, adjusting the
thermostat is equally dominant as adjusting the local cooler (32.14%). Under the hot condition,
drinking cold beverage both in the physical and virtual offices follows adjusting local heater.
Interestingly, under the hot condition in the virtual office, interaction with the desk fan has been
observed as much as drinking cold beverage (21.43%), while it is not the type of interaction that
is chosen (7.14%) in the physical office. One potential reason could be the idea of interacting with
a virtual red fan, which has simulated motion effects was more interesting, as opposed to the no

75
visual simulations like in thermostat, local heater/cooler, or radiant heater. Under the cold
condition, a difference occurred in adjusting the thermostat (3.57% in physical, 32.14% in virtual).
A potential reason could be the ease of thermostat control in the virtual office. For participants to
adjust the thermostat in the virtual office, it would suffice to get close to the thermostat with the
controller. However, in the physical office, it might have been harder for them to lean forward to
reach the thermostat to adjust. In addition, under the cold conditions, adjusting the desk fan and
the clothing level were not among the first interactions, thus they were not the emergent adaptive
interactions. We provided the desk fan option in the cold conditions to be able to level the number
and type of interactions. However, it was already cold, and the participants did not need any extra
ventilation, at least not as an emergent adaptive interaction. The reason for not observing the
clothing level as the first interaction could be related to the climate. Participants were already
wearing warm clothes and it is not customary to have coats, etc. in Mediterranean climates, like in
Los Angeles; thus, the participants did not have any extra clothing to put on.
6.5. Limitations and Future Studies
The present study shows promising results for using virtual reality in the thermoception domain,
as well as enriching virtual realism in occupant behavior studies by adding thermal stimuli to the
experience. Yet, using virtual environments has some limitations. Due to motion sickness,
longitudinal studies cannot be conducted using virtual environments. In addition, since the haptics
component of immersion was not included in this study, there was a slight difference between
virtual and physical offices. Participants could perceive the tactile surface pressure when
interacting with the heating/cooling remedies and feel cold/hot sensations as they got close to these
remedies. These sensations did not translate to the virtual environment as participants were using
a controller. In this study, we focused on indoor air temperature due to the climatic characteristics

76
of Los Angeles. Due to its Mediterranean climate, humidity was not an influential factor. A
previous study highlighted that the assumption of indoor air temperature is equal to mean radiant
temperature could be made in large spaces without radiating sources [284]. Another study showed
that the difference between indoor air temperature and mean radiant temperature is negligible
unless there are large window areas in the environment. They concluded that this difference is
negligible when there is no or a small amount of direct solar radiation is present in the room [285].
We conducted our study in a large windowless office. Thus, similar to other studies [286–289], we
assumed the mean radiant temperature equal to the indoor air temperature. Future studies could
incorporate other thermal characteristics (e.g., air velocity, mean radiant temperature) of indoor
environments for thermal comfort estimations.
We conducted this study in a contextually enriched virtual office in order to improve the realism
of the environment by providing interactivity between participants and the objects in the
environment. Future studies could try to understand the impacts of virtually created
objects/settings on occupants’ thermoception; model and compare different virtual environments
with different contextual settings. In this study, we used an abstraction of the physical office as an
interactive virtual office – in other words the virtual office was identical to the physical office in
terms of its layout, size, furniture, colors, interactive options and so on. Future studies could study
the level of detail needed in virtual environments to adequately represent the physical
environments. The diversity of experimental tests could also be improved by testing our
hypotheses with the same methodology in different virtual environments. Another future approach
could be testing these hypotheses in a virtual environment and having participants experience this
environment in a different room than where they experience the physical environment sessions.

77
Our long-term objective is to understand human-building interactions by using the information
extracted from various human subject experiments through perceptual decision-making processes.
Thus, we plan to conduct experiments in virtual environments to test different scenarios, such as
changing building orientation, office layouts and/or interior designs, lighting configurations to
understand the influence of contextual variables on occupant responses to built environments.
6.6. Conclusions
In this study, we compared virtual environments to physical environments with respect to the
influence of thermal stimuli on the selected response variables (i.e., actual versus perceived indoor
air temperature, thermal comfort and satisfaction, number and type of interactions), and addressed
the Research Question 1.1 and Objective 1. This comparison showed that there exists no
significant difference between virtual and physical environments with respect to thermal comfort
and satisfaction. Thus, subjective thermal and satisfaction assessment methods could be used as
markers of thermoception in virtual environments by measuring the perceived thermal comfort
and satisfaction in these environments. We also analyzed the participants’ ability to perceive
indoor air temperature: a hybrid metric of subjective (perceived indoor air temperature) and
objective feedback (sensor data of indoor air temperature). Participants were not able to
perceptually quantify the indoor air temperature neither in physical nor virtual environments. In
this respect, we found no significant difference between the two environments. We tested
participants’ virtual environment experiences through presence and immersive tendencies and
analyzed their effect on other response variables. The results showed successful immersion in
virtual offices, with lower presence in the cold condition. One important lesson learned is that
presence is an important factor to monitor and control in behavioral studies in virtual
environments, since participant involvement and interactions are directly related to the feeling of

78
being present [290]. There exists (marginally) no significant difference in the number of
interactions between virtual and physical environments. The analyses on the type of first
(emergent) interactions showed that in both environments, participants’ initial adaptive
interactions are similar. Thus, the type of interactions could be used as markers of thermoception
in virtual environments.

79
Chapter 7. Understanding Human-Building Interactions Under Multimodal
Discomfort in Single Occupancy Offices
Occupants’ interactions with building systems, and occupant-related factors influence building
energy consumption [4]. The uncertainty of occupant behavior limits the accuracy of building
performance simulations during the design phase, and as a result occupants’ comfort and
satisfaction during the operation phase of the building life cycle. Thus, it is imperative to
understand occupants’ behavior (defined as any response of an individual or groups to their
environment [5]) within built environments. Yet, a holistic approach for quantifying HBIs (e.g.,
trends, patterns, interaction probabilities) with heating/cooling remedies and lighting fixtures are
missing in the current state of art. Although individual interactions are hard to predict, interaction
related trends and patterns for groups of building occupants could be retrieved from observation
studies [16] and could potentially provide useful insights regarding HBIs (Human Building
Interactions). Thus, we aim to retrieve the information regarding HBIs through quantifiable means
and metrics (i.e., trends, patterns, probabilities, number, type, kind of HBIs).
In this study, we focused on co-presence of visual and thermal discomfort (i.e., multimodal
discomfort). In mundane office contexts, multiple interaction means (e.g., blinds, thermostat),
multiple cues (i.e., visual and thermal) co-exist. Thus, the motivation is to inform future realistic
occupant behavior models and building performance simulations for more accurate energy
consumption predictions and improved occupant satisfaction with built environments. The reason
for focusing on these two stimuli (visual and thermal) is the large energy consumption share of
heating/cooling remedies and lighting systems. HBIs with heating/cooling remedies and
lighting/daylighting fixtures were often studied in siloes (e.g., [24, 25]). Providing the co-presence

80
of these two stimuli could help us to understand the occupants’ interactions relatable to buildings’
thermal and visual system operations.
We embodied the high-resolution HBIs through decisions and responses, derived metrics
attributed to interactions, such as number, type and occurrence probability of these interactions.
Our unique contribution is providing the missing high-resolution information (e.g., patterns,
probabilities) regarding occupant interactions. By this way, future studies could also leverage this
rich occupant information and inform the future human-centered design and renovation decisions.
Our research objective is to understand the decision-making processes to create an insight on how
occupants respond to multimodal discomfort. We devised a human subject experiment and
conducted it in an Immersive Virtual Environment (IVE). IVEs provide a multimodal sensory
experience (i.e., visual, haptic, auditory, olfactory, thermal and gustatory or any combination of
them) to users and provide an enhanced control over the extraneous variables [40] and enable
experimenters to conduct controlled experiments. In our previous study, we showed the adequacy
of adding ambient thermal cues to IVEs in HBIs context [291]. Thus, we used IVEs to perceptually
envelop the participants in controlled virtual experiment settings and provide multimodal cues.
In this study, we focused on perceptual decision-making for understanding HBIs under multimodal
discomfort (i.e., solar gain and synthetic reflected disability glare in our context). Decisions take
place in a built environment, thus understanding how occupants respond to environmental stimuli
would help us understanding HBIs. These responses could be physiological, psychological, or
physical (which is mediated by the former two types of responses). In our study, we only focused
on the physical interactions, which are mediated by the physiological and psychological responses.
We collected participant data regarding energy consumption related interactions (which are simple
adjustments providing rapid ambient changes [23]; adjusting the blind, ceiling lamps, task lamps

81
[292], thermostat, fan and a heater) under multimodal discomfort stimuli to understand HBIs
through perceptual decision-making. The following sections explain our experimental study and
provide a discussion of results.
7.1. Design of Experiment
The experiment setting was an office space of thirteen square meters (Figure 5). Each participant
was welcomed to take a seat on the chair in front of the window, facing the door, his/her back
turned to the window (Figure 5). We designed this sitting configuration to avoid outside view as a
confounding variable. Each participant was recruited to either multimodal discomfort in south-
facing office (group 1) - or no discomfort in north-facing office (group 2) session (randomly
assigned). All participants were immersed in the same virtual environment, only the orientation
and its dependent room visual characteristics (i.e., daylight and synthetic glare) and thermal
characteristics (i.e., solar gain related temperature variations) were different. Participants were
asked to perform a generic office task (i.e., reading and comprehension tasks) as in mundane office
environments. We did not conduct any task performance assessment; the tasks were for keeping
the participants focused in the virtual environment.

Figure 5. Experiment office space (initial conditions) (top: north-facing office; bottom: south-facing
office)

82
Participants were required to complete surveys before and after the experiment session. We
recorded all occupant-building system interactions throughout the experiment. During the
information session, participants were informed that they would be able to adjust the visual and
thermal characteristics of the environment and perceive the potential changes as a result of their
interactions. Visual stimuli were conveyed in the virtual environment; thermal stimuli were
provided in the physical environment through real heating/cooling remedies upon participants’
interactions with the controls in the virtual environment. Participants were wearing head-mounted
displays. Thus, they did not see the experimenter changing the thermal environment by adjusting
heating/cooling remedies upon their interactions. However, they were able to perceive the thermal
change as an output of their interactions.
7.2. Environment and Apparatus
The experiment workstation included a computer monitor, a keyboard, a mouse, a desk fan, a
radiant heater, a thermostat, a task lamp, and a ceiling light switch to control the ceiling lamps
from the workstation. We modeled and rendered the virtual office in Rhino, 3DS Max, and Unity
3D. Interaction options (e.g., turning on the desk fan, choosing ventilation level 1 to 3) were
provided through C# scripts in Unity. An Oculus Rift head-mounted display was used for
immersing the participants in this virtual office. We simulated multimodal sensory discomfort (i.e.,
solar gain as a thermal and synthetic disability glare as a visual discomfort factor) in the discomfort
condition by leaving the blind half-open. Meaning, the participants were welcomed to an office
with the blind half open and with solar gain as a consequence. Thus, if participants were assigned
to this condition, they experienced higher room temperature in the room, and synthetic glare on
their monitor. We simulated realistic daylight cycle using Unity engine and added synthetic glare
using “Amplified Bloom” asset in Unity. By this way, synthetic glare was computed in accordance

83
with simulated daylight cycle and programmed to be sensitive to lighting changes (i.e., simulated
artificial and daylighting changes, such as changes in blind position) in the office. A previous study
estimated the peak indoor solar gain as 2.7 C degrees in south-facing spaces [293]. In our study,
we a had half-closed blind, thus we assumed the initial indoor solar gain as approximately 2-3 C
degrees in the discomfort condition, and no solar gain in the no discomfort condition. To isolate
the perceived multimodal sensory discomfort, we provided the same initial configuration (i.e., all
heating/cooling remedies off, task lamps and ceiling lamps off, and blind half open) in both
conditions. We kept the visual configuration (i.e., office layout, interior design, furniture,
heating/cooling remedies, lighting fixtures, outside view, work station orientation) and
participants’ activity levels (i.e., office tasks) the same in both conditions. We recorded indoor air
temperature and humidity by SMAKN DHT22 AM2302 digital temperature and humidity
measurement sensor to control the initial thermal environment (with temperature accuracy of +-
0.5C and humidity accuracy of +-2% RH (Max +-5%RH)).
7.3. Experiment Details
7.3.1. Recruitment
Prior to the recruitment, the Institutional Review Board (IRB) of University of Southern California
approved the study. We recruited 90 participants (45 in Group 1 – multimodal discomfort condition
in a simulated south-facing office, 45 in Group 2 – no discomfort condition in a simulated north-
facing office). The participants were recruited through the USC’s human subject pool (who
received course credits for their participation), or they were volunteers. We determined motion
sickness either through self-reporting, or the simulator sickness surveys. If anyone experienced
motion sickness anytime, we thanked and dismissed them from the study.

84
7.3.2. Pre-experiment session
We held the information session in a physical room (different than the experiment room) with
neutral indoor air temperature (i.e., 22-23 C). During this session, participants read the IRB-
approved informed consent form. After the information session, participants completed the pre-
experiment survey. This survey was designed to retrieve information regarding participant
demographics. They were also required to complete the simulator sickness questionnaire. We used
the 15-20 minutes of pre-experiment session as an acclimatization period to the neutral air
temperature for decreasing the influence of their prior thermal state. Then, we explained different
interaction options to participants, and how to use the controller for their interactions. Afterwards,
we took the participants to another room where they were welcomed to the physical experiment
setting (Figure 6).

Figure 6. A participant immersed in the experiment setting through Oculus Rift HMD and controlling
the environment through a controller
7.3.3. Experiment session
The experiment session was conducted in two groups: Group (1) multimodal discomfort condition
and Group (2) no discomfort condition. No discomfort condition had neutral thermal environment

85
(22 to 23 C indoor temperature) and a lit environment without any visual discomfort. Multimodal
discomfort condition had solar gain (26 to 27 C indoor temperature, slightly warmer thermal
environment as the thermal discomfort factor) and synthetic glare (visual discomfort factor). These
non-extreme perceivable multimodal sensory discomfort factors were provided by simulating the
same virtual office by changing the building orientation (north versus south). These were the initial
conditions, and participants had control over the thermal and visual characteristics of the
environments through heating/cooling remedies (i.e., a desk fan, a thermostat, a radiant heater)
and lighting fixtures (i.e., a task lamp, a pair of ceiling lamps each with 3 light bulbs, a blind)
(Figure 7). If the participants decided to interact with the heating/cooling remedies and the lighting
fixtures, they were provided with the perceptual realism of experiencing the outcome of their
interactions. For example, if a participant turned on the desk fan and changed the speed of the fan
in the virtual environment, he/she perceived the ventilation outcome of his/her interaction. As
another example, if a participant turned on the ceiling lamp, he/she perceived the increased lighting
intensity in the environment compared to its previous state. Thermal and visual virtual realism
through these interactions were in accordance with the current state of art [110, 111, 291].

Figure 7. Interaction means: (1) desk fan, (2) thermostat, (3) radiant heater, (4) task lamp, (5) ceiling
lamp, (6) blind

86
We randomly assigned the participants to group 1 or group 2, and we also randomized the
execution order of group 1 and group 2 experiments to avoid the influence of outdoor temperature
on the participants’ responses. In both conditions, we provided the same number and type of
interactions. Also, at all times, the number of heating/cooling remedies and lighting/day lighting
fixtures were equal. Each remedy and fixture had different levels of control. Heating/cooling
remedies interaction options were: (1) adjusting the desk fan (with 3 speed levels), (2) adjusting
the radiant heater (on/off), (3) adjusting the thermostat (on/off). Lighting/daylighting fixtures
interactions options were: (1) adjusting the task lamp (4 lighting intensity levels), (2) adjusting the
ceiling lamps (3 lighting intensity levels), (3) adjusting the blind (up and down movement of blind
with 25% increments, which provided 5 different levels; closed, and 25%, 50%, 75%, 100% open).
The ‘on’ setting of the thermostat had 23 C degrees setpoint, and it operated as a cooling mean.
Participants were informed that the thermostat is to turn on/off the air conditioning system.
Interactions took place through a controller with pre-programmed interaction buttons. When
participants interacted with heating/cooling remedies, the experimenter simultaneously
implemented the change in the physical environment, so the participants would perceive the
simulated realistic outcome of their thermal interactions. We also simulated the local sensation of
solar gain by placing weak electric heating panels behind the participants. Likewise, participants
interacted with the lighting/daylighting fixtures through the controller, but they perceived the
change through lighting intensity change in the virtual office.
Participants were instructed to complete the four reading comprehension tasks. Meanwhile, if they
decided to adjust the environment, they could perform those interactions. Participants spent
approximately 15 minutes either in the no discomfort or multimodal discomfort condition. During
this period, they were free to adjust the environment as much as they wanted and whenever they

87
wanted. We measured HBI parameters: number, type, hierarchical order, time (i.e., response time),
patterns of interactions, and the immersion experience (i.e., presence and immersive tendencies)
both in the north- and south-facing virtual offices.
7.3.4. Post-experiment session
Subsequent to completion of the experiment session, participants were welcomed back to the room
where the information session took place and completed the post-experiment surveys in this room.
The post-experiment surveys were designed to retrieve the information regarding participant’s
interactions, visual and thermal comfort and satisfaction, presence, immersive tendencies,
simulation sickness. These-surveys included the Slater-Usoh presence questionnaire to measure
the participants’ presence in virtual offices, Immersive Tendencies Questionnaires (ITQ) to
measure their tendency of presence in virtual environments. After completing this session, the
participants were thanked and dismissed.
7.3.5. Data analysis
After data cleaning, we imported and organized our data, and performed statistical analyses on
samples using Statistical Package for Social Sciences (SPSS) version 23. We reorganized the data
depending on the type of statistical analysis (e.g., Mann Whitney, Chi Square, Loglinear tests).
SPSS output included information regarding the study samples and p/U values (i.e., probability of
occurrence of an event representing the level of significance in hypothesis tests). We performed
post hoc tests on our data set and found that majority of our sample data were not normally
distributed. Thus, we focused on conducting non-parametric statistical analyses techniques (i.e.,
Mann Whitney Tests) for our hypothesis tests. If the significance values (i.e., p-values) are below
0.05, there are significant differences between the two samples, and we reject the null hypotheses.

88
7.4. Results and Discussion
To ensure the sample homogeneity between the two conditions, we hypothesized that the
participants in both samples have similar visual (i.e., lighting) and thermal environment
preferences, presence, and demographics (i.e., age, gender, ethnicity, education level). We did not
find any difference regarding age, ethnicity, and education level yet, we found significant
difference between samples with regards to gender (i.e., p=0.001<0.05, number of females in
Group 2 were more than in Group 1). We did not find any significant difference in visual
(p=0.158>0.05) and thermal (p=0.161>0.05) environment preferences between the samples. Our
results show that the majority of the participants in both of the conditions prefer a combination of
natural and artificial lighting, and prefer neutral thermal environments (i.e., 21.6 to 23 C) (Table
4). In addition, we compared the two conditions with regards to participants’ presence to ensure
presence, or lack thereof does not influence the results. We did not find any significant difference
in presence (0.478>0.05) between the participants in the two conditions.

89

Table 4. Visual and thermal environment preferences of participants in two conditions
7.4.1. Perceptual decisions
7.4.1.1. Type of decisions
We analyzed the type (i.e., thermal or visual) and kind of immediate decisions, reflecting their first
response to the environment. We conducted a series of Chi-Square tests and found that the two
conditions were significantly different with regards to the type of immediate decisions
(p=0.03<0.05). Although in both conditions the majority of the immediate decisions were visual,
variations in the number of these decisions between these conditions (i.e., 23 vs 33, Table 5) caused
this significant difference. In our post-experiment surveys, we asked the participants to evaluate
the perceived synthetic glare on their work plane prior to and after their interactions. We compared
the perceived synthetic glare in the two conditions, using Mann Whitney Tests. Perceived synthetic
glare prior to interactions was significantly different in Group 1 than Group 2 (p=0.0001<0.05),
however, it was similar (p=0.961>0.05) in Group 1 and 2 after the interactions which shows the
No Discomfort
(Group 2)
Multimodal Discomfort
(Group 1)
# of
participants
%
participants
within
sample
# of
participants
%
participants
within
sample
Visual
environment
preference
Daylighting 19 42.2 12 26.7
Artificial lighting 1 2.2 0 0
Combination of natural
and artificial lighting
25 55.6 33 73.3
Thermal
environment
preference
Cool thermal condition
(20-21.1 C)
8 17.8 15 33.3
Neutral thermal
condition (21.6-23 C)
22 48.9 21 46.7
Warm thermal
condition (23.3-24.4 C)
15 33.3 9 20

90
dominance of visual decisions’ over thermal decisions was related to synthetic glare. In Group 1,
this reflects the immediate response to the multimodal discomfort; in Group 2, it potentially
reflects the first preferred change in the environment. Also, the dominance of visual decisions in
Group 2 could be related to the slower thermal versus rapid visual environment response to
participants’ interactions. Thermal discomfort takes longer (cooling takes time) than visual
discomfort to remove from the environment.
Conditions Type of Decision Number of Participants % participants in the sample
No
discomfort
(Group 2)
Thermal 18 4.4
Visual 23 20
No decisions to
interact
4 25.6
Multimodal
discomfort
(Group 1)
Thermal 12 13.3
Visual 33 36.7
No decision to
interact
0 0

Table 5. Type of immediate decisions in the two conditions
After understanding the type of immediate decisions with respect to the presence of discomfort,
we analyzed the kind of immediate decisions (e.g., blind, thermostat, desk fan) and level of
adjustments (e.g., blind 25% open, desk fan speed level 1, thermostat on). We performed Chi
Square tests and found that kind of immediate decisions were significantly different
(p=0.001<0.05) in the no discomfort condition and multimodal discomfort condition (Table 6).
These findings show the majority of the participants in the multimodal discomfort condition tried
to restore their visual comfort (through interactions with blind) and thermal comfort (through
interactions with thermostat and fan) in response to the multimodal discomfort. Immediately
deciding to adjust the blind dominates the other interaction means. This is consistent with our
previous findings regarding to the type of immediate decisions; the majority of the participants in
the multimodal discomfort condition had immediate visual adjustment decisions. As all

91
participants were immersed in mundane office realism, they were asked to complete a generic
office task. Thus, Due to the synthetic glare on the monitor, the majority of the participants in the
multimodal discomfort condition tried avoiding the synthetic glare by adjusting the blind to
proceed with the office task. Adjusting the blind also provided a local thermal relief (i.e., due to
the disappearance of solar radiation on participants’ backs) in the shorth term and an ambient
thermal change in the long term. Thus, interactions with the blind could be interpreted as an
interface between visual and thermal discomfort. Another important finding is that in the
multimodal discomfort condition, the second and third dominant decisions were adjusting the
thermostat and desk fan; 15.6% and 11.1% of the participants, respectively. These findings show
that the participants focused on restoring their thermal comfort by adjusting the thermostat and the
desk fan. In the multimodal discomfort condition, everyone had at least one decision to adjust the
environment. However, due to lacking discomfort factors, four (8.9%) participants in the no
discomfort condition decided to keep the initial visual and thermal environments, forty-one
participants decided to adjust the environment. Adjusting the desk fan was the most frequent
decision in Group 2, overall the visual decisions were more dominant. Based on the most frequent
immediate decisions, participants were more focused on the visual discomfort (i.e., ~58% adjusted
the blinds) in Group 1, whereas participants were more focused on the thermal discomfort (i.e.,
~27% adjusted the desk fan) in Group 2.

92
Decision No Discomfort
(Group 2)
Multimodal Discomfort
(Group 1)
Number of
participants
% participants
within sample
Number of
participants
% participants
within sample
Keep initial
setting
4 8.9 0 0
Adjust desk fan 12 26.7 5 11.1
Adjust
thermostat
2 4.4 7 15.6
Adjust radiant
heater
4 8.9 0 0
Adjust blind 10 22.2 26 57.8
Adjust task
lamp
3 6.7 2 4.4
Adjust ceiling
lamp
10 22.2 5 11.1

Table 6. Decisions in two conditions
7.4.1.2. Number of decisions
We performed Mann Whitney Tests and did not find any significant difference between the number
of decisions in the two conditions (p=0.358>0.05). However, the number of decisions in the no
discomfort condition ( North=4.6) were more than in the multimodal discomfort condition
( South=2.9) (Figure 8). The maximum number of decisions in the no discomfort condition was
fifteen, while it was eight in the multimodal discomfort condition. Likewise, there exists no
significant difference between the two conditions with regards to number of thermal
(p=0.118>0.05) and visual (p=0.256>0.05) decisions, separately. Similarly, the number of thermal
decisions ( Group 2=1.78> Group 1=1.20) and visual decisions ( Group 2=5.59> Group 1=4.64) in the no
discomfort condition was more than in the multimodal discomfort condition. As shown in Figure
8, we observed a J-curve shaped decreasing trend in the number of participants as the decision
index increases. It is worth mentioning that the decision index was the ordered number of decisions
that enabled us to see how many participants had decisions per index (e.g., as shown in Figure 8a,

93
41 participants in no discomfort condition had first decisions (index 1)). We also observed a higher
decision rank in Group 2 than Group 1 at all times; the absence of discomfort in Group 2 potentially
made the participants act based on their visual and thermal preferences rather than the urge to
restore their comfort as in the multimodal discomfort condition (Group 1).

(a) Thermal and visual
decisions
(b) Thermal decisions (c) Visual decisions

Figure 8. Number of decision-makers versus decision number index (a) thermal and visual, (b) thermal,
(c) visual decisions
Although we did not find a significant difference between the two conditions with regards to the
number of decisions, we observed a decreasing number of decisions trend at all times. Under
multimodal sensory discomfort, participants tend to have less decisions (both visual and thermal)
than no discomfort conditions. Additionally, the number of decision-makers follow a J-curve
shaped decrease with increasing number of decisions (thermal, visual, and both types of decisions).
7.4.1.3. Occurrence probabilities of decisions
Previous studies [14, 26, 104] highlighted the need for granular information regarding HBIs for
understanding occupant behavior. By providing multiple lighting fixtures and heating/cooling
remedies to interact with, we were able to derive the probabilities of each interaction decision. As
in Figures 9 and 10, we calculated and plotted the probabilities of decisions in both conditions. It
is worth mentioning that we kept the detailed levels of the same interaction mean (e.g., desk fan

94
had four adjustment levels; three speed levels and off, we calculated the decision occurrence
probabilities of all four levels). As shown in Figures 9 and 10, the occurrence probabilities of
decisions vary in the two conditions. Due to lacking discomfort in Group 2, probabilities vary more
than in Group 1. The decisions in Group 2 indicate the diversity of thermal and visual preferences,
not immediate response to restore comfort. The most frequent immediate decisions in Group 2
were adjusting the desk fan to speed level 1 (20%) and the ceiling lamp to level 3 (16%).

Figure 9. Detailed decision probabilities in the no discomfort condition (different tones of the same color
refer to the same remedy)
Conversely, in Group 1, participants experienced multimodal sensory discomfort. We can see their
immediate need to retain comfort in occurrence probabilities of decisions (Figure 10). The most
frequent immediate decision was setting the blind to level 1 (25% open) (40%), and the thermostat
to level 1 (turning on the thermostat) (13%). These occurrence probabilities can potentially be used

95
in probabilistic occupant behavior models and integrated to energy simulations by future research
efforts.

Figure 10. Detailed decision probabilities in the multimodal discomfort decision (different tones of the
same color refer to the same remedy)
7.4.1.4. Response time
We also recorded the decision/response time (i.e., time each decision took place) in minutes. We
focused on comparing the decisions made in no discomfort condition to multimodal discomfort
condition with regards to the immediate decision time, and the response/decision rate. We defined
the response rate as a simplified average rate representing how fast the participants responded from
the beginning to the end of the experiment:

𝑑𝑒𝑐𝑖𝑠𝑖𝑜𝑛 𝑟𝑎𝑡𝑒 =
|𝑓𝑖𝑛𝑎𝑙 𝑑𝑒𝑐𝑖𝑠𝑖𝑜𝑛 𝑡𝑖𝑚𝑒 − 𝑖𝑚𝑚𝑒𝑑𝑖𝑎𝑡𝑒 𝑑𝑒𝑐𝑖𝑠𝑖𝑜𝑛 𝑡𝑖𝑚𝑒 |
𝛴𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑑𝑒𝑐𝑖𝑠𝑖𝑜𝑛𝑠

We compared the two conditions with regards to the immediate and final decision times using
Mann Whitney Tests. We did not find any significant differences (Table 7). One key finding is

96
that the mean response time in Group 1 was higher than in Group 2. Faster responses in Group 1
reflect the emerging need to respond to multimodal discomfort (Table 7). We also compared the
two conditions with regards to thermal, and visual, and all (i.e., thermal and visual together)
decisions using Mann Whitney Tests. We did not find any significant difference between the two
conditions (i.e., pT=0.118, pV=0.256>0.05, pTV=0.123 respectively). Despite the similarity of the
two conditions with regards to decision rate, participants decided to interact slower in the no
discomfort condition ( Group 2,TV=0.73) than in the multimodal discomfort condition ( Group
1,TV=1.22). This was potentially due to the lacking discomfort factor in Group 2 which resulted in
a more natural decision scheme and pace of setting the environment to the preferred conditions.
Conversely, in Group 1, the emerging behavior towards restoring comfort fastened the decisions.

Immediate response time
(min)
Final response time (min) Response rate
Therm
al &
visual
Thermal
only
Visual
only
Thermal
& visual
Thermal
only
Visual
only
Thermal
& visual
Thermal
only
Visual
only
p 0.086 0.208 0.140 0.748 0.347 0.188 0.089 0.304 0.946
Group
2
3.18 3.56 3.18 6.71 4.91 5.59 0.73 0.70 0.69
Group
1
2.71 2.53 2.33 6.38 4.11 4.64 1.22 0.67 0.76

Table 7. Means and mean comparisons of immediate and final response times and the response rate
7.4.1.5. Decision patterns
We also investigated the patterns regarding the participants’ decisions in the two conditions. To
easily recognize the patterns, we converted our detailed decisions data (i.e., with all levels of
adjustments) to categorical data. We used data visualization techniques for decisions pattern
recognition, specifically Sunburst diagrams (from Java Script-D3 library) (Figure 11), which

97
enabled us to detect the patterns with regards to type and hierarchical order of the decisions. The
rings represent the hierarchical order of the decisions. Meaning, the first inner ring is the first
decision, the next outer ring is the second decision, and so on. Each colorful segment in the rings
represent a particular kind (e.g., turning on the desk fan) of decision taking place in that decision
hierarchy.

Figure 11. Decision patterns two conditions
([https://gokceozcelik.github.io/Sunburst/c-04/TVGroup%202DecisionCateg/index.html],
[https://gokceozcelik.github.io/Sunburst/c-04/TVGroup%201DecisionCateg/index.html],
[https://gokceozcelik.github.io/Sunburst/c-04/], [https://doi.org/10.5281/zenodo.2483288] [294])

Focusing on thermal and visual decisions together (Figure 11) provides a holistic understanding
of decision patterns. We identified dominant and sub-dominant decision patterns. We assumed that

98
if 20 or more participants had the same initial decision and 6 or more of them had the same second
decision, that pattern was ‘dominant’; likewise, we assumed that if 10 or more participants had the
same first decision and 3 or more of them had the same second decision, that pattern was ‘sub-
dominant’. Consistently, we could not determine any dominant patterns in Group 2: there existed
two sub-dominant patterns (the first decision being desk fan on (24.4%) and the second decision
being ceiling lamp on (6.67%) or the task lamp on (6.67%)) but no sub-dominant patterns existed
after the second decisions.
In Group 1, every participant had at least one decision since participants were more focused on
restoring their comfort due to multimodal discomfort stimuli. We found a dominant and a sub-
dominant decision pattern in Group 1. First adjusting blinds to less than half open (44.4% of the
participants), then turning on the desk fan (22.2% of the participants) are the two dominant
decision patterns in Group 1. These patterns confirm the presence of a large group of participants’
need to avoid synthetic glare and potentially solar gain by setting the blinds to less than half open
(since blind operation also influences the ambient room temperature), then to avoid the solar gain
by turning on the desk fan. Turning on the thermostat (13.3% of the participants), then setting the
blinds to less than half open (8.89% of the participants) was the sub-dominant pattern in Group 1.
This sub-pattern has an opposite priority order compared to the dominant pattern (avoiding thermal
discomfort is prioritized) and potentially a similar reason to the dominant pattern; avoiding
multimodal discomfort. Despite the potential similarity of reasoning, we observed a separation of
the thermal decisions; turning on the desk fan (22.2%) (first thermal decision in the dominant
pattern) versus turning on the thermostat (13.3%) (first thermal decision in the sub-dominant
pattern). In our experiment, we simulated realistic local solar gain in front of the window using an
electric panel heater. This local sensation might have triggered participants to find local solutions

99
(i.e., desk fan) instead of changing the ambient thermal environment (using thermostat). We did
not find any dominant patterns after the second decisions.
For further inferences on the effect of one discomfort factor on the other, we performed Loglinear
analysis using subjective visual and thermal comfort and satisfaction votes. We assumed that the
initial (i.e., prior to initial decision) comfort and satisfaction are potential triggers for the
interaction decisions. Our previous findings showed the dominance of visual decisions over
thermal. Thus, we focused on the effect of initial visual comfort and satisfaction for resampling
our data to perform Loglinear analysis. We chose twenty participants (i.e., ten initially visually
uncomfortable and ten comfortable) in Group 2, likewise in Group 1. Then, we performed
Loglinear analysis to test the potential effects of initial visual comfort on final (i.e., after the last
decision) visual comfort and, initial and final thermal comfort. We found a significant interaction
effect (p<0.0001) between initial visual and thermal comfort. This indicated the significant effect
of initial visual comfort on HBI decisions at different values of initial thermal comfort. We
performed a similar analysis using subjective satisfaction votes. In this analysis, we chose twenty
participants (i.e., ten initially visually satisfied, ten dissatisfied) from Group 1 and 2. Consistently,
we found a significant interaction effect (p<0.0001) between initial visual and thermal satisfaction.
This also indicated the significant effect of initial visual satisfaction on HBI decisions at different
values of initial thermal satisfaction. We also found significant interaction effects between initial
visual satisfaction and final thermal satisfaction (p=0.004), and initial thermal and final visual
satisfaction (p=0.046). These results confirm that there is an effect of one discomfort factor on the
other with regards to initial and final comfort and satisfaction votes.

100
7.5. Limitations and Future Studies
The findings of this study are promising for understanding the HBIs in offices through occupants’
decisions based on their perceptual decisions. Pursuing a systematic assessment of occupant
response metrics enabled us to understand the decision variations between no discomfort and
multimodal discomfort conditions. By using virtual environments as an experimental tool, we were
able to conduct a controlled behavioral study and immerse participants in a contextually rich office
environment wherein they experienced the mundane realism. Despite these benefits, only short-
period empirical studies are feasible using virtual reality due to health-related adverse effects of
VR tools (i.e., simulator-sickness). However, our experiment design framework could potentially
be adopted in future longitudinal occupant behavior studies in physical test-beds. It is worth
mentioning that in our study we created a realistic abstraction of daylight, and likewise synthetic
disability glare to provide visual stimuli to the participants. Although our synthetic glare was
realistic based on the daylight cycle and provided the amplified occlusion on the virtual monitor,
there exist may factors influencing the real glare. A study on virtual visual prototyping [295] tested
the perceived brightness and realism of synthetic disability glare and found that synthetic glare
improves the perceived brightness, yet could only show the brightness match of synthetic glare to
real glare for the high intensity levels of the luminaire. Although we cannot make a generalizable
conclusion, we assumed that high illuminance-caused synthetic disability glare has brightness
match with reality. Future studies could elaborate on synthetic glare by further exploring its
brightness match to reality in different contexts. In this study, we designed the room layout in a
way that the participants sat with their backs turned to the window. We used this configuration to
avoid outside view as a confounding variable, so we could isolate glare as a discomfort factor.
Future studies should follow the common practice of having occupants directly or indirectly facing

101
the windows and elaborate on our findings. Our sample was limited to educated young office-
workers, which might have influenced the generalizability of our findings. Additionally, we found
that Group 1 and 2 were different with regards to gender. Thus, following our methodology, future
studies could expand on this sample to contribute the generalizability of HBI findings. We focused
on cognitive processes of choosing from the options based on human senses and responding to the
environment physically through their interactions. Thus, we stratified the environmental realism
by hybridizing the visual power of virtual environments, and sensual realism of physical
environments. Our long-term objective is to understand HBIs focusing on perceptual decision-
making. In this study, we focused on understanding HBIs through perceptual decision-making and
as a result physically responding to the multimodal sensory changes in single occupancy offices.
In our future studies, we will elaborate on our findings to understand HBIs in different contexts
(e.g., multi-occupancy offices).
7.6. Conclusions
In this study, we focused on understanding HBIs through perceptual decision-making processes
under multimodal discomfort. We compared the nature and type of immediate decisions and found
that they are significantly different in the two conditions (no discomfort vs. multimodal
discomfort). Majority of the participants first focused on restoring their visual comfort then their
thermal comfort. One potential reason for this can be our contextual focus on office workers who
had generic office tasks to complete (which require visual engagement). Consistently, the majority
of the participants in the multimodal discomfort condition immediately decided to interact with
the blind, while majority of the participants in the no discomfort condition immediately decided to
adjust the desk fan. Thus, we concluded that the immediate decisions indicate the emerging need
to restore comfort in the discomfort condition and preferred first choices in no discomfort

102
condition. We observed the type and order of decisions in no discomfort condition were more
scattered than in multimodal discomfort condition. Thus, we did not find any dominant decision
pattern in the no discomfort condition. In the multimodal discomfort condition, the dominant
pattern was adjusting the blinds to less than half-open to avoid the synthetic glare, then adjusting
the desk fan to avoid solar gain. With these inferences, we decided to obtain decision information
at a granular level and we calculated the occurrence probabilities of decisions. These probabilities
also confirmed the diversity of decisions in the no discomfort condition. We did not find a
significant difference between the two conditions with regards to the number of decisions.
However, we observed a decreasing trend for the number of decisions. Likewise, we did not find
any significant difference in response time. Yet, the response time in the no discomfort condition
was higher than the multimodal discomfort condition. Due to lacking discomfort, perceptual
decisions in no discomfort condition reflect the participants' visual and thermal preferences. Our
findings agree with the previous studies’ findings regarding the diversity of occupant preferences
[33, 103]. Conversely in the multimodal discomfort condition, we observed the multimodal-
discomfort driven emerging decisions to restore comfort. Type and order of decisions and the
dominant decision pattern reflected the need to avoid the synthetic glare then solar gain; decisions
took place in a shorter time-span and faster in the multimodal discomfort condition. This study
provides the information and insights regarding HBI patterns and occurrence probabilities through
perceptual decisions in sensually realistic, contextually rich built environments. Through this study
we addressed the single occupancy part of research question 2.1 of objective 2.

103
Chapter 8. Understanding Human-Building Interactions Under Multimodal
Discomfort in Multi Occupancy Offices
Our objective is to understand human-building interactions under multimodal discomfort in multi-
occupancy offices. With this specific study outlined here, we answered the multi-occupancy part
of research question 2.1 of objective 2.
8.1. Methodology
We designed and conducted a human subject experiment with 60 participants. 30 participants were
recruited to the single occupancy office (15 participants experienced no discomfort, 15 participants
experienced multimodal discomfort), and 30 were recruited to the multi-occupancy office (15
participants experienced no discomfort, 15 participants experienced multimodal discomfort). We
pursued a between-subject experiment design. Thus, each participant experienced either a single
or multi-occupancy session, either in the no discomfort or multimodal discomfort condition.
Participants were randomly recruited to either a single or multi-occupancy office, and experienced
either no discomfort or multimodal discomfort condition. Our methodology was similar to our
previous study’s methodology (for the design of experiment, environment and apparatus,
experiment detail, please see Chapter 7) with minor changes. Thus, here we only covered the
differences from the previous study’s methodology.
To simulate a multi occupancy office, we modified our virtual model. First, we modeled an avatar
as the participant’s co-worker and placed him at the desk across the participant. The avatar should
be realistic and interactive [238]. Thus, we ensured the avatar looks professional (i.e., business
attire) and realistic (i.e., he had an interactivity feature; he was performing a computer-based
typing task, his hand motion was recognizable by the participant, but not distracting) (Figure 12).
However, we did not enable any communication between the avatar and the participant to avoid

104
any biased decisions. After the information session, only two participants asked if they could talk
to the avatar. Different from our previous study, we verbally informed the participants regarding
having an office mate right before the session.

Figure 12. Single (left) and multi-occupancy (right) office
We collected HBIs data through perceptual decision-making in a multi-occupancy office. We used
the following quantifiable metrics: number, type, kind, response time as attributes of interaction
decisions. To understand the participants’ responses to the multi-occupancy context, we asked
them to complete multi-occupancy surveys before (i.e., social avoidance and distress evaluation)
and after (i.e., co-presence, altered participant behavior, perceived agent awareness questionnaires)
the experiment session. These surveys helped us understand the participants’ psychological
responses to the virtual office-mate and provided us a background on how multi-occupancy was
perceived by the participants. We posited the virtual office-mate as a stimulus that might trigger
the human-building interaction decisions. The social avoidance and distress survey is a 28-item
self-rated survey measuring social anxiety (i.e., distress, discomfort, fear, anxiety, social
avoidance) [238]. Co-presence is the extent to which an individual has the feeling of being with
someone else [238]. The altered participant behavior survey measures the extent to which
participants alter their behavior in response to agents. We also conducted a perceived agent
awareness survey. This survey measures the extent to which the participants perceive the virtual
office-mate (i.e., agent) as being aware of them [238]. Additionally, we collected data using post

105
experiment surveys regarding the reasons of participants’ interaction decisions. We grouped these
reasons as (1) reported reasons (i.e., participants reported the reasons of their decisions in a post-
experiment survey), (2) underlying reasons (i.e., internal or external determinants of the interaction
behavior). Reported reasons could be: (1) discomfort, (2) personal history, (3) preferences, (4)
discomfort/dissatisfaction with the outcome of previous interaction, (5) ease of interaction, (6)
playing with the environment, (7) no reason/no decision. Underlying reasons of decisions include
age, gender, education level, duration lived in a location and so on. A full list of underlying reasons
(i.e., internal and external determinants of occupants’ energy consumption behavior) could be
found in Chapter 9, Figure 14. In this study, we covered the HBI decisions and their reasons in
multi-occupancy context. It is worth mentioning that the interaction percentages regarding the
reasons were different than our analyses on immediate decision types and kinds. This is because
as a part of our experiment design, participants were given the freedom to report multiple reasons
for each interaction, and also some did not report reasons for their interactions.
8.1.1. Data analysis
After data cleaning, we organized our data. Then, we imported it to the Statistical Package for
Social Sciences (SPSS) version 23 for performing statistical analyses on single and multi-
occupancy samples. We started our analyses with the multi-occupancy surveys to understand how
multi-occupancy (i.e., virtual office-mate) was perceived by the participants. Then, we performed
different types of statistical analyses (i.e., MANOVA, Chi Square tests, Mann Whitney Tests,
Pearson Correlation) and we reorganized the sample data per analysis method. We focused on the
hypotheses tests for understanding the similarities and differences between the HBI decisions in
single and multi-occupancy offices under multimodal discomfort. If the significance levels (i.e.,
p-values and Mann Whitney U-values) are below 0.05, there exist statistically significant

106
differences between the two samples, and we reject the null hypothesis. Lastly, because multi-
occupancy surveys provide a background but not detailed causes for participants’ HBI decisions,
we focused on the reported and underlying reasons of the decisions.
8.2. Results and Discussion
To ensure sample homogeneity between single and multi-occupancy conditions, we compared the
participants in both samples with regards to visual and thermal indoor environment preferences,
presence levels, demographics (i.e., age, gender, ethnicity, education level). We did not find any
difference regarding these measured metrics with respect to the discomfort level (i.e., no- and
multimodal discomfort) and occupancy level (i.e., single and multi). The majority of participants
preferred a combination of natural and artificial lighting. Yet, in the no discomfort and single
occupancy conditions, daylighting and combination of daylighting with artificial lighting were
equally preferred (i.e., 46.7%). The majority of participants preferred neutral thermal
environments (i.e., 21.6 to 23 C) (Table 8). We also compared the single and multi-occupancy
samples with regards to perceived presence; we did not find any low presence issues and did not
find a significant difference (p=0.676) between the samples.

107

Table 8. Visual and thermal environment preferences of participants in all conditions
8.2.1. Multi-occupancy survey responses
We found that only one participant in the no discomfort condition had high social avoidance and
distress tendency and the majority had intermediate (normal) social avoidance and distress
tendency (Figure 13). Thus, the majority of participants reported being comfortable in social
contexts and interactions. We performed a Mann Whitney test and did not find a significant
difference between the no discomfort and multimodal discomfort conditions (p=0.081) in terms of
social avoidance and distress.
% participants
within sample

No Discomfort
(Group 2)
Multimodal Discomfort
(Group 1)

Single
Occupancy
Multi
Occupancy
Single
Occupancy
Multi
Occupancy

Visual
environment
preference
Daylighting 46.7 26.7 40 13.37
Artificial
lighting
6.7 6.7 0 6.67
Combination
of natural
and artificial
lighting
46.7 66.7 60 80
Thermal
environment
preference
Cool
thermal
condition
(20-21.1 C)
26.7 33.3 26.7 13.3
Neutral
thermal
condition
(21.6-23 C)
40 40 40 60
Warm
thermal
condition
(23.3-24.4
C)
33.3 26.7 33.3 26.7

108

Figure 13. Social avoidance and distress with respect to discomfort level
We did not find a significant difference (p=0.934) between co-presence in different discomfort
levels. Yet, in the multimodal discomfort condition, mean co-presence was slightly higher than in
the no discomfort condition. This might imply a higher awareness of an office-mate in the
multimodal discomfort condition wherein multiple stimuli evoked the HBI decisions.
The mean participant responses to the altered participant behavior survey were aligned with the
co-presence survey; participants tended to alter their behavior slightly more in the multimodal
discomfort context than in the no discomfort context. Likewise, our Pearson Correlation tests
showed that there was a significant correlation between co-presence and altered participant
behavior under multimodal discomfort (p=0.046, r=0.522). However, we did not find a significant
difference (p=0.935) between the no discomfort and multimodal discomfort conditions with
regards to altered participant behavior. We found that on average, participants did not make much
extra physical effort (i.e., interactions) to avoid disturbing their office mates.
We did not find any significant difference (p=0.982) in perceived agent awareness with respect to
discomfort levels. Yet, we found that there was a slightly higher perceived agent awareness in the
multimodal discomfort condition. This could be related to higher alertness in the multimodal

109
discomfort condition as a way of responding to the environment in order to restore comfort.
Overall, the mean perceived agent awareness was lower than other multi-occupancy survey
metrics. This is normal since the virtual office mate was programmed to have realistic body
motions while focusing on his work (i.e., typing on his laptop) and was not designed to respond to
the participants’ queries or interactions. Similarly, the experimenter did not observe any participant
querying the virtual office-mate. We focused on the potential psychological impact (e.g., peer
pressure) of merely the presence of an office mate on occupants’ multimodal HBIs rather than the
communication and action-reaction (i.e., physical response of the virtual agent to the changes in
the office as a result of participants interactions with lighting fixtures and heating/cooling
remedies) between the participants and the virtual office mate.
According to some of the survey questions, the majority of participants reported that the virtual
office mate was not aware of them and participants did not feel observed by it. Previous studies
found a relation between the gaze behavior of virtual agents and the participant’s personal space
[296]. Thus, adding gaze behavior to the virtual office-mate would add an extra stimulus to the
experiment which would potentially become a confounding variable. Thus, we eliminated the gaze
behavior which might have influenced the perceived agent awareness. Finally, we conducted a
series of Pearson Correlation tests and found that the perceived agent awareness significantly
increased with higher initial response time (p=0.05, r=0.514), initial (p<0.0001, r=0.862) and final
(p=0.004, r=0.700) thermal response time, and initial visual response time (p=0.043, r=0.528) in
the no discomfort condition. We concluded that perceived agent awareness in the no discomfort
condition caused participants to decide slower to interact. This delay was especially effective on
thermal decisions, given both initial and final response times were influenced.

110
8.2.2. Perceptual decisions
We analyzed the type (i.e., visual or thermal), kind (e.g., desk fan, task lamp), number and response
time of immediate decisions as they define the emerging occupant response to the environment.
8.2.2.1. Type of decisions
We performed MANOVA (multivariate analysis on variance) and Chi Square tests and did not
find any significant difference between the type of immediate decisions with regards to occupancy
level (p no discomfort=0.741; p multimodal discomfort =0.283). Yet, we found that the visual decisions were
more frequent than the thermal decisions (Table 9). Immediate thermal adjustments took place
less frequently in the multi occupancy office than in the single occupancy office, both in the no
discomfort and multimodal discomfort conditions. Conversely, immediate visual decisions took
place more frequently in the multi-occupancy office than in the single occupancy office, both under
the no discomfort and multimodal discomfort conditions. Given the presence of multimodal
discomfort, and given the participants were required to complete a generic office task on their
monitor, they prioritized restoring their visual comfort over thermal comfort. It could also be
related to the visual decisions having an immediate indoor environmental change while thermal
decisions taking longer in terms of the environmental change. Having a virtual office mate might
have accelerated their decision-making, amplified the discomfort awareness and made them
prioritize the visual decisions for completing the generic office tasks.

111

Table 9. Type of immediate decisions
Participants under multimodal discomfort potentially prioritized avoiding the glare and/or solar
gain. We compared the perceived synthetic glare on the work plane prior to and after their
interaction decisions. We did not find any significant differences in perceived glare with respect
to the occupancy level. In both occupancy levels, the majority of participants did not perceive glare
in the no discomfort condition prior to and after their interaction decisions. On the other hand, the
majority of participants perceived the synthetic glare in the multimodal discomfort condition prior
to their decisions and did not perceive it subsequent to their decisions (they restored their comfort).
Thus, the decisions under multimodal discomfort could potentially be related to the synthetic glare
regardless of what the occupancy level of the office was.
Additionally, having visual interactions dominate thermal interactions also might have had
underlying reasons. Therefore, we analyzed the underlying reasons of HBI decisions with Chi
Square tests with regards to the discomfort level (i.e., no discomfort versus multimodal
discomfort). We found that the perceived glare prior to decisions and discomfort level were
significantly associated (p=0.001). Initial perceived glare in the multimodal discomfort condition
(73% of the multimodal discomfort sample) was more than the no-discomfort condition (32%).
We did not identify a significant difference between the no discomfort and multimodal discomfort
% participants within sample

Comfort Level

Occupancy Level

No
Discomfort
Multimodal
Discomfort

Single Multi Single Multi
Type of decisions Thermal 20 13.3 20 6.7
Visual 60 73.3 80 93.3
No decision 20 13.3 0 0

112
conditions with regards to perceived visual comfort prior to decisions. Yet, we found that thermal
satisfaction prior to immediate decisions was significantly associated with the discomfort level
(p=0.009). The majority of participants in the no discomfort condition (60%) were satisfied with
the initial thermal environment, yet, half of the participants in the multimodal discomfort condition
(50%) was dissatisfied with it. We concluded that the multimodal discomfort was thoroughly
perceived in the multimodal discomfort condition both in single and multi-occupancy offices.
Thus, participants performed interactions to avoid discomfort in the multimodal discomfort
condition whether they had an office mate or not, and the presence of glare was the main reason
for visual interactions.
Then, we focused on the underlying reasons for decisions with regards to the occupancy level (i.e.,
single versus multi-occupancy). We found significant associations (p=0.025) between the
occupancy level and the duration lived in Los Angeles. Our results showed that the majority of
participants in single-occupancy office lived in Los Angeles longer than the ones in multi-
occupancy office. These differences might be part of the reasons of our results regarding HBIs in
multi-occupancy offices. The longer participants lived in the same region/climate, the more
adapted they were to the thermal characteristics of that region [155, 297]. The difference between
the single and multi-occupancy samples in terms of the duration participants lived in Los Angeles
might have influenced their thermal decisions (i.e., they were able to tolerate heat) and might have
amplified their visual decisions compared to thermal decisions under multimodal discomfort
condition. Participants that lived in Los Angeles longer than the others might be more tolerant to
non-extreme warmth in their environments due to mild climate.
8.2.2.2. Kind of decisions
We continued our analyses with the kind (e.g., blind, ceiling lamp) of immediate decisions. We
performed MANOVA and Chi Square Tests on our sample, and we did not find any significant

113
difference between the kind of immediate decisions in single and multi-occupancy offices
(pmultimodal discomfort=0.741, pno discomfort=0.962). Both under no discomfort and multimodal
discomfort conditions, the majority of participants prioritized interacting with the ceiling lamp and
the blind. In the no discomfort condition, the majority of participants decided to interact with the
ceiling lamp both in single (33.3 %) and multi (33.3%) occupancy offices. In the no-discomfort
condition, the second most dominant immediate decision was adjusting the blind (i.e., 20% and
26.7% of the participants in single and multi-occupancy offices, respectively). Under the
multimodal discomfort condition, the most dominant immediate decision was adjusting the blind
(60%), and the second most dominant decision was adjusting the ceiling lamp (~27%) in the multi-
occupancy office. Results were similar in the single occupancy office (~47% of the participants
interacted with the blind, ~27% with the ceiling lamp). 20% and 13.3% of the participants in single
and multi-occupancy offices in the no discomfort condition respectively kept the initial settings.
Participants in the multimodal discomfort had at least one interaction decision regardless of the
occupancy level.
Having multimodal discomfort, the majority of participants tended to avoid glare and mostly
interacted less with the lighting fixtures (e.g., 60% adjusted the blind). They interacted only with
the thermostat, and the number of participants who adjusted the thermostat was approximately
three times more in the single occupancy office under multimodal discomfort (to avoid the solar
gain) than the other three conditions (Table 10). Interacting with the thermostat less and the blind
more under the multimodal discomfort condition in the multi-occupancy office might also signify
a psychological response of tolerating the solar gain in the short term (i.e., adjusting the blind
would also help with avoiding the solar gain in the longer term) in order not to disturb the virtual
office mate with interactions, yet, avoiding glare to perform generic office tasks.

114

Table 10. Kind of decisions
We performed Chi Square tests with regards to the immediate decision kinds for understanding
the underlying reasons of HBI decisions. We found significant associations between the kinds of
decisions and visual comfort (p<0.0001), visual satisfaction (p=0.001) and thermal satisfaction
(p=0.017) prior to initial decisions (Table 11). Due to visual discomfort prior to their interactions,
35% of the (60) participants interacted with the blind, 25% with the ceiling lamps. Also, due to
visual dissatisfaction 27% of the participants interacted with the blind, 20% with the ceiling lamps.
The blind was an interaction mean with both thermal and visual outcomes. We also found that
13% of the participants interacted with the blind and they were dissatisfied with the initial thermal
comfort of the experiment setting. Almost 2% of the participants interacted with the thermostat
due to thermal dissatisfaction, 5% interacted with the desk fan because they were neither satisfied
nor dissatisfied, and no one immediately interacted with the radiant heater.

% participants within sample

Comfort Level

Occupancy Level

No Discomfort Multimodal Discomfort

Single Multi Single Multi
Kind of
decisions
Keep initial setting 20 13.3 0 0
Adjust desk fan 13.3 6.7 0 0
Adjust thermostat 6.7 6.7 20 6.7
Adjust radiant
heater
0 0 0 0
Adjust blind 20 26.7 46.7 60
Adjust task lamp 6.7 13.3 6.7 6.7
Adjust ceiling lamp 33.3 33.3 26.7 26.7

115

Table 11. Underlying reasons significantly associated with decisions in multi occupancy office

We also performed multiple Chi Square tests and Loglinear analyses to elaborate on our findings
regarding decision kinds and to understand the reported reasons of visual interactions being more
dominant than thermal. As a result of our Loglinear analyses, we found a significant association
(p=0.012) between the discomfort level (i.e., no discomfort and multimodal discomfort) and the
immediate decision kind. As a result of the Chi Square analyses, we found that discomfort was the
most dominant reported reason both under the no discomfort (56.7%) and multimodal (63%)
discomfort conditions regardless of the occupancy level. We found a significant association
(p=0.012) between the immediate decision kind and discomfort (as a reported reason). The
majority of decisions due to discomfort in the no discomfort condition were adjusting the ceiling
lamps (30% of the no-discomfort sample) and adjusting the blind (17%). Conversely, the majority
of discomfort related decisions under the multimodal discomfort condition was adjusting the blind
(37%) and adjusting the ceiling lamps (17%). This change in decision kind dominance with respect
Kind of Immediate Decisions (% of the sample of 60
participants)

Determi
nant
Category No
decision
Desk
Fan
Therm
ostat
Radiant
Heater
Blind Task
Lamp
Ceiling
Lamp
To
tal
Prior
visual
comfort

Uncomfortable 0 1.7 3.3 0 35 6.7 25 71.7
Comfortable 8.3 3.3 6.7 0 3.3 1.7 5 28.3
Prior
visual
satisfacti
on
Dissatisfied 0 1.7 0 0 26.7 3.3 20 51.7
Neither dissat.
Nor sat.
0 1.7 0 0 3.3 3.3 5 13.3
Satisfied 8.3 1.7 10 0 8.3 1.7 5 35
Prior
thermal
satisfacti
on
Dissatisfied 0 0 1.7 0 13.3 5 11.7 31.7
Neither dissat.
Nor sat.
0 5 1.7 0 5 1.7 6.7 20.1
Satisfied 8.3 0 6.7 0 20 1.7 11.7 48.4

116
to discomfort level (e.g., 30% of the participants in no discomfort condition interacted with the
ceiling lamps, whereas 17% of the multimodal discomfort sample interacted with it) explains the
significant difference of decision kinds we identified in our Loglinear analyses. Presence of
multimodal discomfort, or lack thereof, influenced immediate decision kinds. Hence, participants
interacted with the blind to avoid discomfort under multimodal discomfort. Yet, their interactions
with the ceiling lamps were more dominant than their interactions with the blind under no
discomfort. To increase the power of our analyses, we performed Chi Square tests on our entire
multi-occupancy sample data for understanding the associations between the reported reasons and
the kind of immediate decisions. We found that discomfort was significantly associated (p=0.026)
with the decision kinds. Adjusting the blind (27%) and the ceiling lamps (23%) due to discomfort
took place more than the other decision kinds (Table 12). Interactions due to discomfort were
mostly visual, thus, this also explains the dominance of visual decisions over thermal decisions
which we highlighted in the previous section.

Table 12. Reported reasons significantly associated with HBI decisions in multi-occupancy office
8.2.2.3. Number of decisions
We performed a comparative analysis on the number of decisions within the first minute of the
experiment. We focused on the first minute because it represents the immediate response of
participants to the presence/lack of an office-mate. This way, we could understand whether the
participants had an initial psychological hesitance (due to the presence of a virtual office-mate)
influencing their physical interactions. We performed MANOVA on the number of decisions
Kind of Decisions (% of the sample of 60 participants)
Reported Reason No
decision
Desk
Fan
Thermos
tat
Radiant
Heater
Blind Task
Lamp
Ceiling
Lamp
Total
No reason/decision 8.3 0 1.7 0 6.7 1.7 0 18.4
Discomfort 0 1.7 5 0 26.7 3.3 23.3 60

117
within the first minute of exposure to the office environment. We did not find any significant
difference between the number of decisions within the first minute with regards to occupancy level.
Yet, we found that the mean number of decisions in single and multi-occupancy offices under the
no discomfort condition were almost equal, but, the number of decisions in single occupancy office
was higher than the multi occupancy office under multimodal discomfort.
We also performed MANOVA and found a significant difference (p=0.014) between the
occupancy levels with regards to the number of decisions. The number of decisions in the multi-
occupancy office ( NoDiscomfort, multi=3.8> NoDiscomfort, single=2.4; NoDiscomfort, multi=5.4> NoDiscomfort,
single=3.7) was more than in the single-occupancy office both under the no discomfort and
multimodal discomfort conditions. We further investigated the potential reasons of this significant
difference and found that the number of visual decisions was significantly different between the
two occupancy levels (p=0.003). The majority of decisions under all conditions was visual. Thus,
the dominance of visual decisions over thermal decisions determined the level of significance of
number of decisions with regards to the occupancy level. We also identified a slightly higher co-
presence under the multimodal discomfort condition than in the no discomfort condition. This
might have caused the lower number of immediate thermal decisions in the multi-occupancy office
compared to the single occupancy office under multimodal discomfort. Overall, the number of
decisions were skewed towards visual decisions due to the factors explained in the previous
sections. Additionally, in the multi-occupancy office, the number of decisions was higher than in
single occupancy offices. This is potentially due to participants being comfortable in social
contexts and interactions, and multimodal discomfort amplified this difference in multi-occupancy
office.

118

Table 13. Number of decisions
We found a significant association (p<0.0001) between the decisions due to reported personal
history and discomfort level. Under multimodal discomfort, the number of decisions due to
personal history in the single occupancy office (30% of the multimodal discomfort sample) was
higher than in the multi-occupancy office (0%). Overall (i.e., regardless of the discomfort level),
the number of immediate decisions due to personal history was significantly associated (p=0.003)
with occupancy level. Immediate decisions due to personal history in the single occupancy office
(25%) were more than in the multi-occupancy office (2%). Thus, we concluded that the personal
history did not play a role regarding HBIs in the multi-occupancy office. This was potentially
related to the peer pressure of having a virtual office-mate. Their decisions were mostly mediated
by other reasons (e.g., discomfort as highlighted in the previous section).
8.2.2.4. Response time
We also recorded the response time (i.e., time corresponding to each decision) of the decisions in
minutes. We focused on comparing the decisions in the single occupancy and multi-occupancy
offices with regards to the immediate and final decision times, and the response decision rate. We
Number of decisions within sample

Comfort Level

Occupancy Level

No Discomfort Multimodal
Discomfort

Single Multi Single Multi
Number of Decisions Minimum 0 0 1 1
Maximum 9 7 7 9
Mean 2.4 3.8 3.7 5.4
Number of visual decisions Mean 1.4 2.8 2.5 3.5
Number of thermal decisions Mean 1 1 1.3 1.9
Number of decisions within the
first minute
Mean 0.47 0.47 0.87 0.60

119
defined the response rate as the average rate representing how fast the participants responded to
the environment from the beginning to the end of the experiment:

𝑑𝑒𝑐𝑖𝑠𝑖𝑜𝑛 𝑟𝑎𝑡𝑒 =
|𝑓𝑖𝑛𝑎𝑙 𝑑𝑒𝑐𝑖𝑠𝑖𝑜𝑛 𝑡𝑖𝑚𝑒 − 𝑖𝑚𝑚𝑒𝑑𝑖𝑎𝑡𝑒 𝑑𝑒𝑐𝑖𝑠𝑖𝑜𝑛 𝑡𝑖𝑚𝑒 |
𝛴𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑑𝑒𝑐𝑖𝑠𝑖𝑜𝑛𝑠

Using MANOVA, we compared the two occupancy levels with regards to the immediate and last
decision times and the response rate. We did not find any significant difference (Table 14) between
the two occupancy levels with regards to the immediate (p=0.238) and final (p=0.494) response
times, and decision rates (p=0.380). However, we found some patterns in the decision response
time analyses. Immediate visual decisions took place faster, the decision-making process lasted
longer (i.e., final response time was higher) and the decision rate was faster in the multi-occupancy
office than in the single occupancy office. This pattern and the overall dominance of the visual
decisions might be related to the participants’ proactive responses to visual discomfort despite
having an office mate. We did not find such patterns for thermal decision response time and rate.
We also found that the immediate decisions took place faster under the multimodal discomfort
condition than under the no discomfort condition. This was due to the presence of discomfort
factors under multimodal discomfort condition which urged the participants to interact with the
environment faster at the beginning of their decision-making process. We also found that perceived
agent awareness in the no discomfort condition caused participants to decide interacting slower.
This delay was especially effective on thermal decisions, given both the initial and final response
times were influenced. Having this delay especially in thermal decisions (i.e., both initial and final
response time) might have contributed to the dominance of visual immediate decisions over
thermal decisions in the no discomfort condition. We did not find similar relations to multimodal
discomfort condition. This lack of correlation, especially regarding the immediate visual decision

120
response time, explains the dominance of visual decisions over thermal. Additionally, lacking this
correlation regarding thermal decision response time denotes the overall amplified response to
multimodal discomfort. These findings also potentially explain the differences between the no
discomfort and multimodal discomfort conditions with regards to number of the decisions within
the first minute patterns (Section 8.2.2.3.).
Immediate response
time (min)
Final response time (min) Response rate
Therma
l &
visual
Ther
mal
only
Visua
l only
Thermal
& visual
Thermal
only
Visual
only
Thermal
& visual
Thermal
only
Visual
only
No
discomfort
Singl
e
2.27 0.93 1.80 3.80 1.67 3.00 0.44 0.52 0.56
multi 1.2 1.40 1.40 5.60 3.00 5.53 1.03 0.7 1.25
Multimodal
discomfort
Singl
e
0.93 2.80 0.93 4.93 4.00 3.33 1.01 0.53 0.76
multi 0.73 2.27 0.80 4.60 3.27 4.13 0.82 0.28 1.02
p 0.238 0.959 0.610 0.494 0.745 0.123 0.380 0.491 0.071

Table 14. Means and mean comparisons of immediate and final response times and the response rate

8.3. Limitations and Future Studies
Our study presents promising results for understanding HBIs in multi-occupancy offices under no
discomfort and multimodal discomfort conditions. Yet, there were some limitations in our study.
We recruited sixty participants to our study; future studies could expand on our findings with a
larger sample size. Another limitation was that the virtual office-mate (i.e., avatar) was realistic
but not communicative. Adding more features (e.g., gaze behavior, communication capabilities)
to the avatar for more realism could create confounding effects that could bias our findings. We
did not observe any participants trying to communicate or interact with the avatar during the
experiment in any means. Thus, we concluded having a realistic abstraction of an office mate (with

121
a suitable formal attire and typing motion) was sufficient to our objectives and did not influence
our results. However, since we contributed to understanding HBIs in the presence of an office
mate, future studies could improve the human-avatar interactions and elaborate on our findings.
8.4. Conclusions
In this study, we focused on understanding HBI decisions in multi-occupancy context. We
performed comparative analyses on quantifiable HBI metrics (i.e., type, kind, number, response
time of decisions). Additionally, we stratified our findings with the reported and underlying
reasons of decisions using post-experiment surveys, and multi-occupancy surveys.
We first focused on multi-occupancy survey responses, which provided us a background on
participants’ perception of social contexts and multi-occupancy, and more insights to our findings
regarding quantifiable HBI metrics. One of the most important findings is that according to the
social avoidance and distress survey, the majority of participants were comfortable with social
contexts and interactions. Thus, participants might have not experienced peer pressure that
significantly influenced their HBI decisions in the multi-occupancy context. We found that the co-
presence, altered participant behavior, perceived agent awareness under multimodal discomfort
were slightly (but not significantly) higher under the multimodal discomfort condition than the no
discomfort condition. We also showed that the increased perceived agent awareness significantly
delayed the initial response time, initial and final thermal response time and initial visual response
time in the no discomfort condition. Yet, lacking these significant relations to perceived agent
awareness under the multimodal discomfort condition signifies the amplified proactive fast-paced
responses to discomfort despite the presence of an office mate.
We did not find a significant difference between HBIs in the single and multi-occupancy offices
with regards to some HBI metrics (type, kind, response time), but we did with regards to the

122
number of decisions. Participants’ comfort with social contexts and interactions, and discomfort
(as a reported reason), caused significantly higher (p=0.014) number of decisions in the multi-
occupancy office than in the single occupancy office. Conversely, the number of interactions due
to personal history was significantly lower in the multi-occupancy office than in the single-
occupancy office under multimodal discomfort. We also focused on the number of decisions
within the first minute of the experiment. We found that the number of decisions was equal in the
single and multi-occupancy offices under no discomfort, and lower in the single than multi-
occupancy office under multimodal discomfort. This was potentially due to the immediate
perception of the virtual office-mate upon initial exposure to the environment. The decision-
making processes changed in the long-term resulting in inferences regarding the number of
decisions being opposite to the inferences regarding the number of decisions within the first
minute. We found that the type of immediate decisions was similar (i.e., not significantly different)
across the two occupancy levels. The immediate visual decisions were more common/frequent
than the thermal. This was due to significant glare perception under multimodal discomfort, and
the longer duration participants in the single-occupancy office lived in Los Angeles. Likewise, the
kind of immediate decisions was also similar between the two occupancy-levels. Yet, we identified
some patterns regarding the immediate decision kinds. We observed interactions with the blind
followed by the ceiling lamp being dominant under multimodal discomfort. This order reversed
under the no discomfort condition. We found that the reported discomfort significantly influenced
the kind of immediate decisions. Immediate response time was also similar between the two
occupancy levels.
Our major contribution is to provide the high resolution HBI information in multi-occupancy
contexts leveraging objective (i.e., number, type, kind, response time of decisions) and subjective

123
(e.g., multi-occupancy surveys, reported reasons) HBI metrics. Additionally, we showed the
importance of incorporating context-specific measures (e.g., social avoidance and distress) in
multi-occupancy surveys to understand the psychological background of HBI decisions. Through
this study we addressed the multi-occupancy part of research question 2.1 of objective 2.

124
Chapter 9. Understanding the Reasons of Human-Building Interactions
Through the investigation explained in this section, we answered the research question 2.2 of
objective 2. As highlighted by domain experts, in the current state of the art, occupant behavior
models integrated in building performance simulations, mostly rely on assumptions regarding
occupant interactions [14, 15]. For example, memoryless models (e.g., Bernoulli process), one of
the frequently used stochastic occupant behavior models (others include Discrete-time Markov
chain, Survival Analysis [14]), assume that an occupant action is only dependent on the previous
action [25]. Thus, these models neglect the possibility of occupants’ behaviors that might have a
previous memory (e.g., residual discomfort from previous interactions, habits, adaptations, and so
on). The reasons for the presence of manual controls (e.g., for window blinds, operable windows,
light switches) in buildings are well-articulated: to avoid discomfort, to accommodate different
space uses (e.g., presentation with a projector in a conference room) and different occupant
preferences [10]. Yet, the reasons of occupants’ interactions with the building controls have not
been explored or assessed systematically (e.g., through a statistical quantification).
A previous study [243] interviewed the occupants of a high-performance building regarding the
reasons of their interactions with the blinds, electric lighting and thermal controls. Some of the
reasons they found were as follows: too cold, too hot, too much glare, to let more light in. Another
interview question was asked to understand the reasons of not having any interactions. The reasons
included social concerns, lack of control, lack of knowledge to use the controls. This conceptual
reasoning provides insight regarding occupants’ decisions to interact or not interact with building
controls. However, currently there exists no quantitative evidence of the reported and underlying
reasons of these interactions.

125
Majority of the current occupant behavior modeling approaches are focused on relations between
interactions and environmental factors [105, 179, 246, 298–302]. Yet, to develop robust occupant
behavior models, the cause and effect relations of HBIs should be incorporated to the knowledge
regarding occupants’ energy consumption behavior [14]. Our objective is to explore the reasons
for occupants’ HBI decisions through the relations between their interactions, potential
determinants of their decisions (i.e., underlying reasons), and the analysis of subjective survey
questionnaires on the reasons of the interactions (i.e., reported reasons). With this, we could
overcome the simplified behavior modeling assumptions in the current state of the art by
contributing to the level of information through adding cause-effect logic to occupant behavior
modelling. The existing model accuracies could rely on occupancy information-based assumptions
if we know how and why occupants interact with their environment. In addition, future studies
could contribute to more robust and stratified occupant models, as suggested by [14], by adding
the related determinants and reasons as model inputs that we provide through our study. With this
study, we aimed to understand the reasons of human-building interactions (HBIs) with a holistic
approach where we concatenated occupancy information across different contexts: single and
multi-occupancy offices. This way, we could also contribute to the generalizability of the reasons
of HBI decision-making.
9.1. Methodology
For addressing the research question 2.2 of objective 2, we aimed to understand the relations
between the decisions to interact and internal and external determinants of these decisions, as well
as the subjective reasoning votes. We pursued the same methodology as our last two studies
(please see Chapter 7 and 8 for the design of experiment, environment and apparatus, experiment

126
details). Here, we describe the differences in this study’s methodology from the previous two
studies (Chapter 7 and 8).
Although the experimental procedure in this study’s methodology was the same as the previous
two, the types of data that we used in the analysis were different than those studies. To explore the
relations of various internal (personal factors such as individual, physical or mental conditions that
influence the human behavior) and external (contextual, physical and system-related factors that
may influence the environment) determinants of the decisions, we measured the potential internal
and external determinants (Figure 14) of occupant behavior and reported reasons (Figure 15) of
the decisions to interact both in single and multi-occupancy offices.

Figure 14. Metrics measured in pre- and post- experiment surveys

127

Figure 15. Reported reasons of interactions measured through post-experiment surveys
These measured metrics had different data types (e.g., numerical, categorical). For the analyses,
our first step was cleaning the data, and performing data type conversions since the HBI data (i.e.,
kind of decisions, discomfort level) to understand the reasons were nominal. After federating the
data types, we performed descriptive statistics, consecutive Chi-Squares Tests to derive
meaningful relational inferences between the kinds of HBI decisions and these metrics. We also
leveraged information visualization techniques (i.e., Sankey diagrams) to better understand the
reasons of interactions by mapping the reasons per decision kind. These analyses enabled us to
understand the reasons of HBI decisions through potential determinants of behavior and subjective
reason reporting. Metrics that were highly correlated to the decisions could be used as key reasons
metrics in future studies. It is worth mentioning that we also compared the reasons in the no
discomfort condition to the multimodal discomfort condition with respect to the identified key
reasons metrics.

128
9.1.1. Data analysis
After data cleaning, we organized our data and concatenated the single and multi-occupancy
experiments’ data (Chapter 7 and 8 data, 129 participants). The reason for having 129 participants
instead of 150 is that we had 21 participants in the single occupancy sessions of the multi-
occupancy experiment that we also used in the single occupancy experiment. To avoid double
counting, we only included the 129 participants’ data in this study. Then, we imported the data to
Statistical Package for Social Sciences (SPSS) version 23 to perform statistical analyses on our
data set. We performed different types of analyses. The majority of our data was nominal, hence,
we performed multiple consecutive Chi Square Tests to understand the reasons for decisions. If
the significance levels (i.e., p-values) are below 0.05, there exist statistically significant association
between the variables.
9.2. Results and Discussion
In our analyses, we leveraged the underlying (i.e., determinants of decisions in Figure 14) and
reported (i.e., post-experiment questions in Figure 15) reasons. Given the high number of potential
reported reasons, for statistical inferences we created seven categories of reasons for both thermal
and visual decisions. This way, we aimed to level the number of categories of reported reasons for
the thermal and visual decisions in our analyses. For categorizing the reported reasons, we referred
to the previous studies [4, 10, 20, 55] that outlined the determinants of occupant behavior.
Attitudinal factors, contextual forces, personal capabilities, habits and routines [20]; trust, personal
values, recent personal experiences [55]; climate, building related characteristics, user-related
characteristics, building systems and their operation, social and economic factors, indoor
environmental quality [4]; discomfort, preferences and different space uses [10] were the
important determinants of occupant behavior. Thus, we covered such determinants in our survey

129
questions, and while creating these seven categories, we ensured categories accounted for these
determinants. These categories are; (1) discomfort, (2) personal history, (3) preferences, (4)
discomfort/dissatisfaction with the outcome of previous interaction, (5) ease of interaction, (6)
playing with the environment, (7) no reason/decision (Figure 15). It is worth mentioning that,
categorizing the listed items in Figure 15 in a leveled fashion across visual and thermal decisions
required us to place these items to the closest listed categories they could belong to. Also, the
participants had the freedom to report multiple reasons for each decision. Hence, this information
lets us leverage the frequencies of the reasons and how each reason could be related to certain
kinds of decisions. Yet, this also means that the percentages regarding kinds of HBIs could be
different than the reported percentages in the previous chapters given each decision could have
multiple reasons.
9.2.1. Underlying reasons of HBIs
To understand the underlying reasons of HBIs that cannot be captured through reported reasons,
we performed a separate analysis to find the associations (i.e., correlations) between the decisions
and their determinants. Since the kind of decisions data was nominal (i.e., categorical), we pursued
consecutive Chi Square tests using the determinants that were nominal. Some of these
determinants were numerical, hence, we performed a categorization on these numerical
determinants (e.g., age, BMI) to federate the data types. Yet, there were some numerical
determinants that could not be possibly categorized (i.e., presence, mood, aesthetics/
attractiveness/visual appeal of the final office lighting configuration). Thus, we did not include
these variables in our analyses. We performed our analyses with respect to the comfort level (i.e.,
no and multimodal discomfort), and the kind of decisions. Participants’ first decisions represent

130
the immediate response to the indoor environmental cues. Thus, in our analysis, we focused on the
participants’ first decisions.
Through Chi Square tests, we found significant associations between the discomfort level (i.e., no
discomfort and multimodal discomfort), and visual satisfaction (p<0.0001), thermal comfort
(p=0.018), thermal satisfaction (p=0.009), perceived glare (p<0.0001) prior to immediate
decisions, caffeine intake in the past 24 hours (p=0.053), duration lived in Los Angeles (p<0.0001),
gender (p=0.018). Visual and thermal dissatisfaction under multimodal discomfort (62% and 44%
of multimodal discomfort sample were dissatisfied, respectively) were higher than under no
discomfort (27% and 21% respectively). The number of participants who were uncomfortable with
the initial thermal characteristics of the office in the no discomfort (19%) condition was lower than
in the multimodal discomfort (38%) condition. We did not find a significant association between
the discomfort level and visual comfort (p=0.577) while there was a significant association with
visual satisfaction (p<0.0001). The majority of participants in both discomfort levels reported that
they were uncomfortable (71% for no discomfort, 76% for multimodal discomfort) with the initial
lighting configuration of the office. This was potentially due to the dominant lighting preferences
for combination of natural and artificial lighting in the no discomfort and multimodal discomfort
samples. The initial lighting configuration of both discomfort levels was natural lighting only. We
concluded that reported comfort, or lack thereof, did not necessarily determine the participants’
visual interaction decisions, but satisfaction was more significant in influencing their decisions.
We also observed that some participants reported that they were comfortable, yet dissatisfied with
the visual and/or thermal characteristics of the office. Hence, their satisfaction was more influential
on decisions than their comfort. This is why although perceived glare was significantly different

131
between the two discomfort levels (80% of the participant in the multimodal discomfort condition
perceived the glare), visual comfort was not significantly different between these levels.
We found that caffeine intake was marginally significantly associated (p=0.053) with the
discomfort level. Caffeine intake in no discomfort sample (i.e., 32% of the participants in this
sample had caffeine intake) was lower than the multimodal discomfort sample (48%). This might
have increased the alertness of the participants in the multimodal discomfort condition and affected
their visual perceptions. We also found that the duration lived in Los Angeles was significantly
associated with (p=0.047) with the discomfort level. The number of participants who lived in Los
Angeles for more than a year in the multimodal discomfort condition (52%) was higher than in the
no discomfort condition (30%). Living longer in a location makes one more accustomed to the
climate, yet, this might be influential on one’s thermal perceptions. Although thermal comfort prior
to interactions was significantly different between the two discomfort levels, 62% of the
participants in the multimodal discomfort condition reported they were comfortable with the warm
initial condition. This might be related to the duration they have lived in Los Angeles and their
adaptation to the warm climate.
We continued our analyses focusing on the underlying reasons of decisions with regards to the
kind of immediate decisions. We found significant associations between the kind of immediate
decisions and visual comfort (p<0.0001) and satisfaction (p=0.017) prior to the initial interaction,
visual satisfaction (p=0.033) posterior to the last interaction, gender (p=0.01). We observed that
the majority of associations were related primarily to the blind, ceiling lamps and desk fan
interactions. 74% of the entire sample experienced visual discomfort prior to their initial
interaction decisions. 36% decided to interact with the blind, 17% with ceiling lamps, 5% with the
task lamp. 45% of the sample reported visual dissatisfaction prior to their immediate interactions.

132
Hence, dissatisfied participants were more than the neither satisfied nor dissatisfied, and satisfied
ones. This was primarily associated with the significance of blind (24%) and ceiling lamp (10%)
interactions due to visual dissatisfaction. Subsequent to all the interactions, 86% of the participants
were visually satisfied, and this satisfaction rate was higher than neither satisfied nor dissatisfied
(8%), and dissatisfied (6%) rates. Therefore, we concluded that the participants were able to
significantly improve their visual satisfaction given they had the freedom to adjust the indoor
lighting characteristics with the provided lighting fixtures. Lastly, we found a significant
association (p=0.018) between gender and the kind of decisions. This is not due to the unbalanced
number of females and males in the sample, given 52% of the sample were female, and 48% were
male, which was not a significant difference. Yet, females interacted with the blind (29% of the
sample) and thermostat (5%) more than males (12%, 4% respectively); males interacted with the
ceiling lamps (14%), desk fan (8%), task lamp (4%) and radiant heater (3%) more than females
(6%, 6%, 3%, 0%, respectively). These results potentially show the gender-specific differences
between the decision-making processes with regards to the environmental changes. Males mostly
interacted with local solutions with faster environmental outcome (e.g., desk fan, task lamp)
through which the perceived change in the visual and thermal environmental characteristics could
be faster. Yet, females mostly interacted with the blind and thermostat which both provide slower
outcomes that take time to change the indoor environment, especially the thermal characteristics
of the room. Although the blind could provide an immediate lighting change in the environment,
thermal changes due to solar gain, and lack thereof, through the blind adjustments were slower
than the kinds of decisions male participants made.

133

Table 15. Underlying reasons significantly associated with decisions in combined sample
9.2.2. Reported reasons of HBIs
Participants’ first decisions represent the immediate response to the indoor environmental cues.
Yet, it is imperative to understand the reasons of these immediate responses to identify the
participants’ needs with respect to these cues. Thus, we performed consecutive Chi-Square tests
to assess the relations of the immediate decisions to the reported reasons. In our analyses, we
focused on the discomfort level (i.e., no discomfort or multimodal discomfort conditions) as it was
the common factor of both the single and multi-occupancy office contexts.
We found that immediate decisions due to playing with the environment (p=0.019) and discomfort
(p=0.003) were significantly associated with the discomfort level. There were more decisions due
to playing with the environment under no discomfort (18%) than multimodal (9%) discomfort.
Also, immediate decisions due to discomfort were more under multimodal discomfort (40%) than
no-discomfort (26%). We found that HBI decisions were significantly associated with reported
Kind of Decisions (% of the sample of 129 participants)
Determinant Category No
decisi
on
Desk
Fan
Thermostat Radiant
Heater
Blind Task
Lamp
Ceiling
Lamp
To
tal
Prior visual
comfort

Uncomfortable 0 10 4.7 1.6 35.7 4.7 17.1 73.8
Comfortable 5.4 3.9 4.7 1.6 5.4 2.3 3.1 26.4
Prior visual
satisfaction
Dissatisfied 0 4.7 1.6 1.6 24 3.1 10.1 45.1
Neither dissat.
Nor sat.
0 1.6 2.3 0 4.7 2.3 3.9 14.8
Satisfied 5.4 7.8 5.4 1.6 12.4 1.6 6.2 40.4
Post visual
satisfaction
Dissatisfied 0 0.8 0.8 1.6 1.6 0 1.6 6.4
Neither dissat.
Nor sat.
0 1.6 0 0 4.7 1.6 0 7.9
Satisfied 5.4 11.6 8.5 1.6 34.9 5.4 18.6 86
Gender Male 3.1 7.8 3.9 3.1 12.4 3.9 14 48.2
Female 2.3 6.2 5.4 0 28.9 3.1 6.2 52.1

134
personal history (p=0.04), preferences (p=0.038) and discomfort (p=0.003). Adjusting the blind
(31%) and ceiling lamps (14.7%) due to discomfort took place more often than the other decision
kinds. We also observed a similar dominant decision kind pattern due to personal history. Yet, we
identified the dominance of adjusting the ceiling lamp decisions was replaced by the desk fan
decision when the reason was preferences. Adjusting the blind (15.5%) and desk fan (7%) due to
preferences were more frequent than the other decision kinds. We concluded that the participants
focus on both visual and thermal characteristics of the environment while setting it to their
preferences. Whereas, the majority of decisions due to personal history and discomfort was visual.

Table 16. Reported reasons significantly associated with decisions in combined sample
Subsequent to our statistical analyses, we also plotted Sankey diagram(s) using all the decisions
for 129 participants under the no discomfort and multimodal discomfort conditions mapping the
kinds of decisions on the reported reasons (Figure 16). There were 477 decisions under the
multimodal discomfort condition and 491 decisions under the no discomfort condition. The most
frequent decisions under multimodal discomfort were setting the blind to 25% open (i.e., Blind 1)
(50% open blind configuration caused glare) and turning on the desk fan to the slowest level (i.e.,
Desk fan 1) (Figure 16). The majority of these most frequent decisions was due to discomfort (i.e.,
reported reason). Meaning, when we focused on all the decisions for all participants, under
multimodal discomfort the most dominant decisions were interacting with the blind and desk fan
to avoid solar gain and glare. Participants also reported that these decisions took place mostly due
Kind of Decisions (% of the sample of 129 participants)
Reported Reason No
decision
Desk
Fan
Thermostat Radiant
Heater
Blind Task
Lamp
Ceiling
Lamp
Tot
al
Personal history 0 4.7 3.9 2.3 7 3.1 5.4 26.4
Preferences 0 7 5.4 2.3 15.5 1.6 4.7 36.5
Discomfort 0 8.5 5.4 0.8 31 4.7 14.7 65.1

135
to discomfort. The second most dominant reason for these two decisions was preferences. Overall,
the majority of decisions was related to discomfort (31%), followed by preferences (17%), playing
with the environment (17%) and personal history (15%).

Figure 16. Sankey visualization of decisions mapped on reported reasons in combined sample
[https://gokceozcelik.github.io/Reasons%20Behind%20HBIs/][303]

To understand the reasons for HBI decisions at a granular level, we also calculated the frequencies
of reported reasons. We found that the most frequently reported reasons for the HBI decisions in
descending order were discomfort, playing with the environment, preferences, personal history,
discomfort or dissatisfaction with outcome of the previous interaction and ease of interaction mean
under the no discomfort condition (Figure 17). It is worth mentioning that some participants
decided to keep the environment as is and some did not report reasons for their decisions. Thus,
we calculated the no reason or no decision frequencies separately. There were many decisions
reported to be related to discomfort in the no discomfort condition. The majority (76%) of these

136
reportings was related to the lighting related interactions. Each decision could have multiple
reported reasons, yet, the lack of discomfort often resulted in playing with the environment and
potentially reporting a perceived discomfort along with it.

Figure 17. Frequency of decisions per reported reason in no discomfort condition
We found that the most frequently reported reasons in the multimodal discomfort condition (Figure
18) were as follows (in descending order): discomfort, preferences, personal history, playing with
the environment, discomfort or dissatisfaction with outcome of the previous interaction, ease of
interaction mean. The frequency of playing with the environment ranked after personal history
under multimodal discomfort, while it was the second most frequent reason under no discomfort.
Meaning, participants prioritized restoring their comfort over playing with the environment under
multimodal discomfort. This also shows that the reported reasons behind HBI decisions under no
discomfort and multimodal discomfort were quite similar except for playing with the environment.

137

Figure 18. Frequency of decisions per reported reason in multimodal discomfort condition
9.3. Limitations and Future Studies
The present study has promising results for understanding the reported and underlying reasons of
HBIs in a representative commercial building with single and multi-occupancy offices. Yet, there
are some noteworthy limitations. We performed our experiments in realistic immersive virtual
abstractions of commercial offices. To avoid potential simulator sickness related problems, we
were limited to keeping our participants in the environment for no more than 20 minutes. Although
the non-extreme conditions (i.e., solar gain, glare, multi-occupancy) were sufficiently conveyed to
the participants in these realistic offices and provided us the occupant related information we aimed
to investigate, future studies could also adopt our methods and techniques for longitudinal studies
if needed. Future studies could leverage our contributions by developing context-based occupant
energy consumption behavior models. Adding the significant reasons we identified as model
predictors could overcome the high-level occupancy assumptions of energy simulations, and
eventually improve the accuracy of energy consumption simulation in commercial settings and
validate the simulation results.

138
9.4. Conclusion
In this chapter, we focused on understanding the underlying and reported reasons of human-
building interactions. We pursued a holistic approach by combining the samples with contextual
variations (i.e., single occupancy, multi occupancy) to obtain more generalizable inferences
regarding the reasons of HBIs. We performed multiple consecutive Chi Square tests focused on
analyses with regards to discomfort level and the kinds of HBI decisions.
We first focused on understanding the underlying reasons of HBI decisions leveraging the
determinants of occupants’ energy consumption behavior. With an increased number of
participants, and accommodating all the contextual variables across the experiments, we obtained
a representative mixed discomfort level and occupancy level sample. As a result of these analyses,
we found that visual satisfaction, thermal state and satisfaction, perceived glare prior to the initial
decision, caffeine intake in the past 24 hours, duration lived in Los Angeles, and gender, all
influenced the HBI decisions with regards to discomfort level (i.e., no discomfort and multimodal
discomfort). We also found that visual state and satisfaction prior to the initial decision, visual
satisfaction posterior to the last decision, and gender were significantly associated with the kind
of immediate decisions. An important finding was regarding the influence of gender on immediate
HBI decision-making. We observed that male participants interacted with local remedies that had
faster visual and thermal environmental outcomes (e.g., desk fan, task lamp). Yet, female
participants interacted with the remedies (i.e., blind and thermostat) that provide ambient and
slower change in thermal environmental characteristics. In this sample, we did not have an
unbalanced gender distribution, hence, we concluded that gender-specific decision-making
schemes could significantly influence perceptual decision-making processes with regards to HBI
decision kinds.

139
Focusing on the reported reasons, we concluded that discomfort, preferences, personal history and
playing with the environment are the most significant reasons influencing the immediate decisions.
We found that the significant reported reasons with regards to discomfort level (i.e., no discomfort
versus multimodal discomfort) were discomfort, personal history and playing with the
environment. Foremost significant reasons determining the immediate decision kinds were
discomfort, preferences and personal history in descending order. Due to discomfort and personal
history, adjusting the blind and ceiling lamps were the most frequent decisions. Yet, HBI decisions
due to preferences replaced ceiling lamp adjustments with desk fan. This signifies that both the
visual and thermal decision types are significantly influenced by preferences, and visual decisions
are more significantly determined by discomfort and personal history than the thermal decisions.
We also captured and visualized the frequency of cumulative HBI decisions (i.e., all the decisions
took place during the experiments) and mapped them on the reported reasons. We found that the
descending frequency order of reported reasons were discomfort, playing with the environment,
preferences, personal history, discomfort/dissatisfaction with outcome of the previous decision, no
reported reasons, and ease of interaction mean under the no discomfort condition. Under
multimodal discomfort, we found that there was a similar accumulated decision-making-reasoning
frequency pattern, except playing with the environment. Due to multimodal discomfort, playing
with the environment was not a priority for the participants, and ranked fourth in the decision
scheme (after personal history).
Our study highlighted the importance of understanding the reasons for HBI decisions by
identifying the underlying reasons based on internal (e.g., caffeine intake in the past 24 hours,
duration lived in Los Angeles) and external (e.g., glare perception) determinants of occupants’
energy consumption behavior, and also the subjective reported reasons. The major contribution of

140
our study is to provide a thorough understanding of the significant reasons and potential
determinants of occupants’ energy consumption behavior with a holistic approach of
accommodating contextual variations (as in real commercial buildings). Through this study we
addressed the research question 2.2 of objective 2.

141
Chapter 10. Limitations and Potential Future Work
10.1. Limitations
Although our studies have promising results for understanding Human-Building Interactions
(HBIs) and their reasons across different contexts (i.e., discomfort and occupancy levels) with an
occupant cognition-centric (i.e., through perceptual decision-making processes) approach, this
work also bears a number of limitations. First, there are some sample-specific limitations. All of
our participants were USC students pursuing undergraduate and graduate level education. They
also were mostly from a similar age range (18-30). They were mostly residing in Los Angeles
(similar location, yet the duration lived in Los Angeles varied as discussed in our results).
Therefore, our results cannot be generalizable to the entire commercial building occupant
population. A larger sample size would be needed for more generalizable results.
Another important limitation that needs to be pointed out is related to the research/experimentation
tool we used. Using virtual environments as an experiment tool enabled us to conduct controlled
behavioral studies and immerse the participants in contextually and sensorially rich office
environments wherein, they experienced the multimodal mundane office realism. We also did not
encounter any problems regarding the samples’ immersion experiences (i.e., immersive tendencies
and presence) except in our thermoception study (Chapter 6), in which the presence was an effect
influenced results in cold versus hot environments. However, due to potential adverse effects of
the simulator (i.e., simulator/motion sickness), participants could not be kept in virtual
environments more than 20 minutes. It is possible to keep participants in virtual environments for
less than 20 minutes and perform experiments with multiple sessions. But longitudinal studies
cannot be conducted in virtual environments. Yet, our experimental design framework could be
adopted by the longitudinal studies and used in physical environments. Another limitation was the

142
feeling of tactile perceptions while interacting with the real remedies (e.g., touching the light
switch to turn on the ceiling lamp) versus lacking the similar tactile perception in virtual
environments (e.g., using the controller to turn on the ceiling lamp). Yet, this was a slight
difference between the virtual and physical environments that did not influence our results given
the multimodal sensory and contextual realism of the virtual environments. It is worth mentioning
that in our studies (Chapters 7, 8, 9), we created a realistic abstraction of daylight and synthetic
disability glare. The synthetic glare was realistic according to the daylight cycle and provided an
amplified disability glare on the virtual monitor enabling an instantaneous visual discomfort
stimulus to the participants. However, there are many factors influencing the real glare. A previous
study [295] testing the perceived disability glare with respect to brightness and realism showed
that the synthetic disability glare improves the brightness, yet the brightness match of synthetic to
real glare only occurred in high intensity illuminances. Thus, future studies should elaborate on
benchmarking the synthetic glare to real glare. Yet, in our studies, participants perceived the
synthetic glare as expected and the complexities of glare did not influence our results. The last
limitation we experienced with virtual environments was regarding the multi-occupancy context
(Chapter 8) and the avatar’s realism as a virtual office-mate. In our study, we created a realistic
abstraction of an office-mate through its formal office attire and hand and body motions engaged
in the typing activity (i.e., a generic office task). However, to avoid potential confounding variables
and related bias, we did not attribute communicative features (e.g., gaze behavior, vocal
communication) to the avatar. Future studies could enhance the avatar realism by adding
communication features to it in order to understand the human-avatar interactions and their
potential influences on HBIs. The realistic abstraction of a virtual office-mate was sufficient to

143
meet our research objectives, yet, adding enhanced communicative features to the avatar might
result in different inferences than ours, given the contextual change.
Another limitation of our studies (Chapters 7, 8, 9) was placing the participants with their back to
the window for capturing their immediate responses to the enhanced perceived glare on their
monitor. Yet, future studies should follow the common practice of having the occupants facing
windows directly or indirectly for glare experience and elaborate on our findings. Lastly, in our
studies, we focused on indoor air temperature to adjust the thermal characteristics of the offices
and we assumed the mean radiant temperature was equal to the indoor air temperature. This
assumption could be made in large spaces without radiant sources [284], or in the environments
with large windows, or when there is no/small amount of direct solar radiation in the room [285].
In Chapter 6, our experiment setting was a large office space with no windows, in Chapters 7, 8,
9 we had a generic window (not large) with the blind half open (either small amount of or no solar
radiation). Thus, we could assume the mean radiant temperature was equal to the indoor air
temperature. Yet, future studies that measure thermal comfort could incorporate other factors (e.g.,
air velocity, mean radiant temperature) influencing the thermal characteristics of the indoor
environment.
10.2. Potential Future Work: Occupant Behavior Modelling and Energy
Consequences of Behaviors
HVAC and lighting systems have a large energy consumption share and there exists an inaccuracy
between the predicted energy performance of buildings and measured actual performance of the
buildings [11, 12] due to the uncertainties of occupant behavior that have not been integrated into
energy simulations [13, 42]. In addition to the energy implications, negligence of occupant
behavior also results in occupant comfort implications. Once occupant behavior is modeled, this
model can be used for developing accurate building energy consumption simulations and

144
strategies. Thus, discrepancies regarding occupant related information in the simulations could be
addressed. Future studies could explore the energy consequences of HBIs by modeling high-
resolution occupant behavior with their reasons and integrating them into building performance
tools. There are two fundamental steps to take for this exploration: occupant behavior modelling
through HBI decisions and constructing a simulation environment.
Occupant behavior can be modeled for future predictions of occupant-system interactions through
their perceptual decisions. Future studies can leverage our findings regarding HBIs and their
reasons, adopt probabilistic (stochastic) occupant behavior modeling methods and translate these
models to a co-simulation environment, wherein the occupant behavior and energy simulation
could run simultaneously and exchange information in real-time for an enhanced occupant
behavior information translation to building performance analysis tools [188], instead of relying
on static schedules for understanding the energy consequences of HBIs. This way, behavior
modeling and simulation efforts based on the high resolution HBI information we provided in this
dissertation can contribute towards understanding the occupants’ influence on building energy
consumption, and (potentially) articulate the discrepancies of existing simulations.

145
Chapter 11. Conclusions
Occupants’ contribution to high energy consumption in commercial buildings (40% of annual
consumption) through HBIs was not assessed in details and it was unknown. This dissertation
explores the high-resolution Human-Building Interactions (HBIs) and their reasons in commercial
settings (i.e., offices) with regards to contextual variations (i.e., discomfort and occupancy levels).
Previous studies highlighted the uncertainties regarding occupant behavior [12, 13] and simplified
occupant-related assumptions [14, 15] as key limitations for accurate building energy consumption
predictions of simulation tools. Thus, in this dissertation, we focused on understanding how and
why HBIs take place in single and multi-occupancy offices to inform the future human-centered
designs and energy analyses by providing this occupant behavior information. To retrieve this
micro-level information, we posited humans as cognitive sensors and used their perceptual
decision-making processes as enablers to understand HBIs. In Chapter 1, we summarized our
studies, highlighted the problems we addressed. In Chapter 2, we explained our motivation for
focusing on understanding HBIs and their reasons across different contexts. In Chapter 3, we
provided a detailed background on occupants’ energy consumption behavior, decision-making
processes, and human response to thermal and visual cues. Chapter 4 explains our detailed
literature review on HBIs (specifically thermoception, occupants’ interactions with blind, lighting
systems, thermostats and heating/cooling remedies) and highlights the limitations and gaps in the
current state of the art. In Chapter 5, we explained our objectives and research questions to address
these gaps and limitations.
In our studies, we emulated the mundane office realism with visual and thermal cues. For the
enhanced realism and controlled experimentation, we used immersive virtual environments as a
research tool in our studies. To date, thermoception in virtual environments in contextually rich

146
spaces was not benchmarked to thermoception in physical environments. In Chapter 6, we focused
on benchmarking thermoception in virtual environments to physical environments to show the
adequacy of using virtual environments, when human thermoception is the foreground in the HBI
context, so that we could add thermal cues to virtual environment in our future studies. Through
this benchmark study, our comparative analyses (i.e., hypotheses tests) showed that there was no
significant difference between virtual and physical environments with respect to thermal comfort
and satisfaction. Thus, we concluded that we could add thermal cues to virtual environments for
enhanced realism and we could also use subjective thermal comfort and satisfaction votes as
markers of thermoception in virtual environments. This study let us create multimodal virtual
environment experiences wherein visual and thermal cues co-existed in the HBI context that we
leveraged in our studies.
In Chapter 7, we focused on understanding HBIs through perceptual decision-making processes
under multimodal discomfort (i.e., solar gain and glare) in single occupancy offices. We found that
under multimodal discomfort the majority of participants focused primarily on restoring their
visual comfort (given the mundane office context required them to perform generic office tasks on
their computers), followed by restoring their thermal comfort. Having multimodal discomfort, the
majority of immediate HBI decisions were adjusting the blind to avoid glare on the monitor. Yet,
under no discomfort, the immediate decisions were mostly adjusting the desk fan. Under no
discomfort, decisions were scattered with no dominant interaction patterns. Yet, under multimodal
discomfort, we found that the dominant interaction pattern was to set the blind to less than half
open (to avoid synthetic glare), then turning on the desk fan to avoid solar gain. Thus, we
concluded that the immediate decisions are indicators of the emerging need to restore comfort
under discomfort, yet, they indicate the preferred first choices under no discomfort. With this

147
study, we provided the high resolution HBI information and insights regarding the patterns and
occurrence probabilities of HBIs through perceptual decisions in sensorially realistic, contextually
rich built environments.
After thoroughly understanding HBIs in single occupancy offices, we focused on understanding
HBIs in multi-occupancy offices in Chapter 8. We did not find significant differences between the
single and multi-occupancy offices with regards to HBI metrics (i.e., type, kind, response time,
number of decisions) except the number of decisions. The number of decisions in the multi-
occupancy office was significantly more than in the single occupancy office. We identified that
this is because the majority of participants were comfortable with social contexts and interactions.
Hence, they potentially did not experience the peer pressure in the presence of a virtual office-
mate. We also found that the visual immediate decisions were more dominant than the thermal
immediate decisions especially in the multi-occupancy office under multimodal discomfort. We
explored the reasons of HBI decisions to explain our findings. We found that under multimodal
discomfort, decisions due to personal history in the single occupancy office were significantly
more frequent than in the multi-occupancy office. We found that the perceived agent awareness
significantly influenced the decision response time in the no discomfort condition. Glare
perception, thermal satisfaction prior to initial decisions, duration lived in Los Angeles
significantly influenced the type of decisions. Visual comfort and satisfaction, thermal satisfaction
prior to initial decisions and reported discomfort significantly influenced the kind of decisions.
Lastly, in Chapter 9, to understand the overarching reported and underlying reasons of HBIs with
a holistic approach of accommodating the contextual variations (i.e., single and multi-occupancy
offices, no and multimodal discomfort) as in real commercial buildings, we combined the samples
experienced these contexts. This approach let us have a larger sample size and more generalizable

148
inferences regarding the reasons for HBIs. The reasons that influenced the HBI decisions with
regards to discomfort level (i.e., no discomfort and multimodal discomfort) were thermal comfort,
visual and thermal satisfaction, glare perception prior to initial decisions, gender, caffeine intake
in the past 24 hours, duration lived in Los Angeles, playing with the environment, personal history
and discomfort. Visual comfort and satisfaction prior to the initial decision, visual satisfaction
posterior to the last decision, gender, personal history, preferences and discomfort significantly
influenced the kind of immediate decisions. Additionally, we obtained the frequencies of reasons
per decision under no- and multimodal discomfort conditions to highlight the dominant reasons of
decisions when the entire decision-making processes for each participant were considered.
Focusing on these frequency schemes, we concluded that the most frequent/dominant reasons of
decisions were discomfort, preferences, personal history and playing with the environment under
both no discomfort and multimodal discomfort. The only difference between the two discomfort
levels with regards to these reasons was playing with the environment which ranked the second
under no discomfort, and fourth under multimodal discomfort.
In Chapter 10, we outlined the limitations of our studies and potential future studies this
dissertation will lead to. Lastly, we summarized the key findings and concluding remarks of our
studies in Chapter 11. The major contribution of this dissertation is to present the HBIs and their
reasons across different contexts (i.e., no- and multimodal discomfort, single and multi-occupancy
offices) using objective and subjective HBI metrics derived from participants’ perceptual decision-
making processes. The findings of this dissertation, representing the statistical inferences regarding
the high-resolution information of occupants’ energy consumption behavior, could enhance the
future human-centered designs, accuracy of energy consumption simulations, sustainable and
occupant-centered energy strategies.

149
References
[1] J.A. Leech, R. Burnett, W. Nelson, S.D. Aaron, M. Raizenne, Outdoor air pollution
epidemiologic studies, Am. J. Respir. Crit. Care Med. 161 (2000) A308.
[2] U.S. DoE, Buildings energy databook, Energy Effic. Renew. Energy Dep. (2011).
[3] U.S. EIA, Annual energy outlook 2013, US Energy Inf. Adm. Washington, DC. (2013).
[4] Z. Yu, B.C.M. Fung, F. Haghighat, H. Yoshino, E. Morofsky, A systematic procedure to
study the influence of occupant behavior on building energy consumption, Energy Build.
43 (2011) 1409–1417.
[5] E. Lindbladh, C.H. Lyttkens, Habit versus choice: the process of decision-making in
health-related behaviour, Soc. Sci. Med. 55 (2002) 451–465.
[6] K.B. Janda, Buildings don’t use energy: people do, Archit. Sci. Rev. 54 (2011) 15–22.
[7] M. Bonte, F. Thellier, B. Lartigue, Impact of occupant’s actions on energy building
performance and thermal sensation, Energy Build. 76 (2014) 219–227.
[8] M. Moezzi, K.B. Janda, From “if only” to “social potential” in schemes to reduce building
energy use, Energy Res. Soc. Sci. 1 (2014) 30–40.
[9] P. Hoes, J.L.M. Hensen, M.G.L.C. Loomans, B. de Vries, D. Bourgeois, User behavior in
whole building simulation, Energy Build. 41 (2009) 295–302.
doi:10.1016/j.enbuild.2008.09.008.
[10] W. O’brien, Occupant-proof buildings: can we design buildings that are robust against
occupant behaviour, 13th Conf. Int. Build. Perform. Simul. Assoc. (2013) 1746–1754.
[11] W. Zeiler, D. Vissers, R. Maaijen, G. Boxem, Occupants’ behavioural impact on energy
consumption: “human-in-the-loop” comfort process control, Archit. Eng. Des. Manag. 10
(2014) 108–130. doi:10.1080/17452007.2013.837252.
[12] A. Mahdavi, E. Kabir, A. Mohammadi, L. Lambeva, C. Pröglhöf, User interactions with
environmental control systems in buildings, in: Proc. PLEA, 2006.
[13] V. Fabi, R.V. Andersen, S.P. Corgnati, F. Venezia, Main physical environmental drivers
of occupant behaviour with regard to space heating energy demand, Heal. Build. 2012.

150
(2012).
[14] D. Yan, W. O’Brien, T. Hong, X. Feng, H. Burak Gunay, F. Tahmasebi, A. Mahdavi,
Occupant behavior modeling for building performance simulation: Current state and
future challenges, Energy Build. 107 (2015) 264–278. doi:10.1016/j.enbuild.2015.08.032.
[15] W. O’Brien, I. Gaetani, S. Gilani, S. Carlucci, P.J. Hoes, J. Hensen, International survey
on current occupant modelling approaches in building performance simulation†, J. Build.
Perform. Simul. 10 (2017) 653–671. doi:10.1080/19401493.2016.1243731.
[16] A. Mahdavi, C. Pröglhöf, User behavior and energy performance in buildings, Wien,
Austria Int. Energiewirtschaftstagung an Der TU Wien. (2009).
[17] P.C. Cacciabue, Guide to applying human factors methods: Human error and accident
management in safety-critical systems, Springer Science & Business Media, 2013.
[18] L. Steg, C. Vlek, Encouraging pro-environmental behaviour: An integrative review and
research agenda, J. Environ. Psychol. 29 (2009) 309–317.
[19] W. Poortinga, L. Steg, C. Vlek, Values, environmental concern, and environmental
behavior a study into household energy use, Environ. Behav. 36 (2004) 70–93.
[20] P.C. Stern, Towards a Coherent Theory of Environmentally Significant Behavior, Journal
of Social Issues, 56, (2000).
[21] G.A. Guagnano, P.C. Stern, T. Dietz, Influences on attitude-behavior relationships a
natural experiment with curbside recycling, Environ. Behav. 27 (1995) 699–718.
[22] C.D. Wickens, J.G. Hollands, S. Banbury, R. Parasuraman, Engineering psychology &
human performance, Psychology Press, 2015.
[23] J. Heerwagen, R. Diamond, Adaptations and Copng: Occupant Response tn Discomfort in
Energy Efficient Buildings, in: ACEEE, 1992.
[24] C. Wang, D. Yan, X. Ren, Modeling individual’s light switching behavior to understand
lighting energy use of office building, Energy Procedia. 88 (2016) 781–787.
[25] F. Haldi, D. Robinson, Interactions with window openings by office occupants, Build.
Environ. 44 (2009) 2378–2395. doi:10.1016/j.buildenv.2009.03.025.

151
[26] C.F. Reinhart, Lightswitch-2002: a model for manual and automated control of electric
lighting and blinds, Sol. Energy. 77 (2004) 15–28.
[27] D. Wyon, Individual control at each workplace for health, comfort and productivity, in:
Proc. Work., 1997.
[28] S.L. City, Preferred surface illuminances and the benefits of individual lighting control : a
pilot study Newsham , G . R .; Arsenault , C .; Veitch , J . A ., (2002) 101–113.
[29] G. Newsham, J. Veitch, C. Arsenault, C. Duval, Effect of dimming control on office
worker satisfaction and performance, in: Proc. IESNA Annu. Conf., Citeseer, 2004: pp.
19–41.
[30] M. Knoop, Dynamic lighting for well-being in work places: Addressing the visual,
emotional and biological aspects of lighting design, in: Proc. 15th Int. Symp. Light. Eng.
Light. Eng. Soc. Slov. Bled, Slov., 2006: pp. 63–74.
[31] H. Juslén, Lighting and Productivity in the Industrial Working Place, Proc. Fifteenth Int.
Symp. Light. Eng. Soc. Slov. Light. Work. (2006) 53–62.
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.114.67&rep=rep1&type=pdf.
[32] J.D. Jennings, F.M. Rubinstein, D. DiBartolomeo, S.L. Blanc, Comparison of control
options in private offices in an advanced lighting controls testbed, J. Illum. Eng. Soc. 29
(2000) 39–60. doi:10.1080/00994480.2000.10748316.
[33] S.A. Sadeghi, P. Karava, I. Konstantzos, A. Tzempelikos, Occupant interactions with
shading and lighting systems using different control interfaces: A pilot field study, Build.
Environ. 97 (2016) 177–195. doi:10.1016/j.buildenv.2015.12.008.
[34] T. Moore, D.J. Carter, A.I. Slater, User attitudes toward occupant controlled office
lighting, Light. Res. Technol. 34 (2002) 207–216.
[35] J. Langevin, P.L. Gurian, J. Wen, Tracking the human-building interaction: A longitudinal
field study of occupant behavior in air-conditioned offices, J. Environ. Psychol. 42 (2015)
94–115. doi:10.1016/j.jenvp.2015.01.007.
[36] M. Indraganti, Behavioural adaptation and the use of environmental controls in summer
for thermal comfort in apartments in India, Energy Build. 42 (2010) 1019–1025.

152
doi:10.1016/j.enbuild.2010.01.014.
[37] Y. Chen, B. Raphael, S.C. Sekhar, Individual control of a personalized ventilation system
integrated with an ambient mixing ventilation system, HVAC&R Res. 18 (2012) 1136–
1152.
[38] J. Kaczmarczyk, A. Melikov, Z. Bolashikov, L. Nikolaev, P.O. Fanger, Human response
to five designs of personalized ventilation, HVAC&R Res. 12 (2006) 367–384.
[39] J. Verhaart, M. Veselý, W. Zeiler, Personal heating: effectiveness and energy use, Build.
Res. Inf. 43 (2015) 346–354.
[40] J. Blascovich, J. Loomis, A.C. Beall, K.R. Swinth, C.L. Hoyt, J.N. Bailenson, Immersive
virtual environment technology as a methodological tool for social psychology, Psychol.
Inq. 13 (2002) 103–124.
[41] E. Delzendeh, S. Wu, A. Lee, Y. Zhou, The impact of occupants’ behaviours on building
energy analysis: A research review, Renew. Sustain. Energy Rev. 80 (2017) 1061–1071.
[42] T. Hong, H. Lin, Occupant Behavior : Impact on Energy Use of Private Offices, Asim
IBSPA Asia Conf. (2012).
[43] H.R. Heekeren, S. Marrett, L.G. Ungerleider, The neural systems that mediate human
perceptual decision making, Nat. Rev. Neurosci. 9 (2008) 467–479.
[44] M.I. Leon, M.N. Shadlen, Exploring the neurophysiology of decisions, Neuron. 21 (1998)
669–672.
[45] F. Haldi, D. Robinson, Modelling occupants’ personal characteristics for thermal comfort
prediction, Int. J. Biometeorol. 55 (2011) 681–694.
[46] O.G. Santin, Behavioural patterns and user profiles related to energy consumption for
heating, Energy Build. 43 (2011) 2662–2672.
[47] A.K. Mishra, M. Ramgopal, Field studies on human thermal comfort—An overview,
Build. Environ. 64 (2013) 94–106.
[48] B. Dong, K.P. Lam, A real-time model predictive control for building heating and cooling
systems based on the occupancy behavior pattern detection and local weather forecasting,
in: Build. Simul., Springer, 2014: pp. 89–106.

153
[49] Z. Yang, B. Becerik-Gerber, The coupled effects of personalized occupancy profile based
HVAC schedules and room reassignment on building energy use, Energy Build. 78 (2014)
113–122.
[50] B. Roisin, M. Bodart, A. Deneyer, P. D’herdt, Lighting energy savings in offices using
different control systems and their real consumption, Energy Build. 40 (2008) 514–523.
[51] J.C. Lam, D.H.W. Li, An analysis of daylighting and solar heat for cooling-dominated
office buildings, Sol. Energy. 65 (1999) 251–262.
[52] O. Sezgen, J.G. Koomey, Interactions between lighting and space conditioning energy use
in US commercial buildings, Energy. 25 (2000) 793–805.
[53] C. Franzetti, G. Fraisse, G. Achard, Influence of the coupling between daylight and
artificial lighting on thermal loads in office buildings, Energy Build. 36 (2004) 117–126.
[54] V. Singhvi, A. Krause, C. Guestrin, J.H. Garrett Jr, H.S. Matthews, Intelligent light
control using sensor networks, in: Proc. 3rd Int. Conf. Embed. Networked Sens. Syst.,
ACM, 2005: pp. 218–229.
[55] P.C. Stern, Energy efficiency in buildings: Behavioral issues, National Academies, 1985.
[56] L.C. Jain, C.P. Lim, Handbook on Decision Making: Vol 1: Techniques and Applications,
Springer Berlin Heidelberg, 2010.
[57] R.A. Wilson, F.C. Keil, The MIT encyclopedia of the cognitive sciences, MIT press,
2001.
[58] P. Slovic, B. Fischhoff, Decision making., (1988).
[59] J. Von Neumann, O. Morgenstern, Theory of games and economic behavior, Princeton
university press, 2007.
[60] K.J. Arrow, Essays in the theory of risk-bearing, (1974).
[61] M. Friedman, L.J. Savage, The utility analysis of choices involving risk, J. Polit. Econ.
(1948) 279–304.
[62] B.A. Mellers, Decision‐Making, Encycl. Cogn. Sci. (1998).
[63] D. Kahneman, P.P. Wakker, R. Sarin, Back to Bentham? Explorations of experienced

154
utility, Q. J. Econ. (1997) 375–405.
[64] S.M. Dillon, Descriptive decision making: Comparing theory with practice, in: Proc. 33rd
ORSNZ Conf. Univ. Auckland, New Zeal., 1998.
[65] G. Klein, Sources of power: How people make decisions, MIT press, 1999.
[66] A. Tversky, D. Kahneman, Judgment under Uncertainty: Heuristics and Biases, Science.
185 (1974) 1124–1131. doi:185/4157/1124 [pii].
[67] C.D. Wickens, J.D. Lee, Y. Liu, S. Gordon-Becker, Introduction to human factors
engineering, (1998).
[68] T. Elliott, Expert decision-making in naturalistic environments: A summary of research,
DEFENCE SCIENCE AND TECHNOLOGY ORGANISATION SALISBURY
(AUSTRALIA) SYSTEMS …, 2005.
[69] F. Alavizadeh, B. Moshiri, C. Lucas, Application of naturalistic decision making to
emergency evacuation simulations, in: Artif. Intell. Theor. Model. Appl., Springer, 2008:
pp. 14–25.
[70] X. Pan, C.S. Han, K. Dauber, K.H. Law, A multi-agent based framework for the
simulation of human and social behaviors during emergency evacuations, Ai Soc. 22
(2007) 113–132.
[71] E.D. Kuligowski, Modeling human behavior during building fires, (2008).
[72] M. Kobes, I. Helsloot, B. De Vries, J.G. Post, Building safety and human behaviour in
fire: A literature review, Fire Saf. J. 45 (2010) 1–11.
[73] J.D. Schall, Neural basis of deciding, choosing and acting, Nat. Rev. Neurosci. 2 (2001)
33–42.
[74] S. Plous, The psychology of judgment and decision making., Mcgraw-Hill Book
Company, 1993.
[75] P.L. Smith, R. Ratcliff, Psychology and neurobiology of simple decisions, Trends
Neurosci. 27 (2004) 161–168.
[76] L. Klotz, Cognitive biases in energy decisions during the planning, design, and

155
construction of commercial buildings in the United States: an analytical framework and
research needs, Energy Effic. 4 (2011) 271–284.
[77] C.J. Hopfe, G.L.M. Augenbroe, J.L.M. Hensen, Multi-criteria decision making under
uncertainty in building performance assessment, Build. Environ. 69 (2013) 81–90.
[78] C.J. Hopfe, Uncertainty and sensitivity analysis in building performance simulation for
decision support and design optimization, PhD Diss., Eindhoven Univ. (2009).
[79] D. Caccavelli, H. Gugerli, TOBUS—a European diagnosis and decision-making tool for
office building upgrading, Energy Build. 34 (2002) 113–119.
[80] M. Jaggs, J. Palmer, Energy performance indoor environmental quality retrofit—a
European diagnosis and decision making method for building refurbishment, Energy
Build. 31 (2000) 97–101.
[81] B.P. Thompson, L.C. Bank, Use of system dynamics as a decision-making tool in building
design and operation, Build. Environ. 45 (2010) 1006–1015.
[82] R. Daft, Organization theory and design, Cengage learning, 2006.
[83] K.H.E. Kroemer, H.J. Kroemer, Engineering physiology: Bases of human
factors/ergonomics, John Wiley & Sons, 1997.
[84] R.W. Bailey, Human performance engineering: Using human factors/ergonomics to
achieve computer system usability, Prentice-Hall, Inc., 1989.
[85] N.W. Heimstra, L.H. McFarling, Environmental psychology, Brooks/Cole Pub. Co., 1978.
[86] J. Malchaire, Methodology of investigation of hot working conditions in the field,
Ergonomics. 38 (1995) 73–85.
[87] B.W. OLESEN, T.L. MADSEN, Measurements of the physical parameters of the thermal
environment, Ergonomics. 38 (1995) 138–153.
[88] R.G. Hoffman, R.S. Pozos, Experimental hypothermia and cold perception., Aviat. Space.
Environ. Med. 60 (1989) 964–969.
[89] V. Candas, A. Dufour, Thermal comfort: multisensory interactions?, J. Physiol.
Anthropol. Appl. Human Sci. 24 (2005) 33–36.

156
[90] L. Kaufman, Color & human response: Aspects of light and color bearing on the reactions
of living things and the welfare of human beings, Psyccritiques. 24 (1979) 532–533.
[91] K. Goldstein, The organism: A holistic approach to biology derived from pathological
data in man., (1939). http://psycnet.apa.org/psycinfo/2004-16223-000 (accessed March 8,
2016).
[92] C. Chan, Does Color Have Weaker Impact on Human Cognition Than Material?, Comput.
Archit. Des. Futur. (. (2007). http://link.springer.com/chapter/10.1007/978-1-4020-6528-
6_28 (accessed March 8, 2016).
[93] A. Yasukouchi, K. Ishibashi, Non-visual effects of the color temperature of fluorescent
lamps on physiological aspects in humans, J. Physiol. Anthropol. Appl. Human Sci. 24
(2005) 41–43.
[94] A. Yasukouchi, The effects of color temperature of light sources on the arousal level and
postural change with different mental tasks, in: Proc Second Int Conf. Human-
Environment Syst. 1998, 1998.
[95] T. Iseki, Effect of color temperature of fluorescent lamps on rectal temperature during and
after heat exposure, in: Proc. 44th Meet. Physiol Anthr. 2000, 2000: pp. 72–73.
[96] C. Cajochen, J. Zeitzer, C. Czeisler, D. Dijk, Dose-response relationship for light intensity
and ocular and electroencephalographic correlates of human alertness, Behav. Brain Res.
(2000). http://www.sciencedirect.com/science/article/pii/S0166432800002369 (accessed
March 9, 2016).
[97] I. Mclntyre, T. Norman, G. Burrows, Human melatonin suppression by light is intensity
dependent, J. Pineal. (1989). http://onlinelibrary.wiley.com/doi/10.1111/j.1600-
079X.1989.tb00412.x/abstract (accessed March 9, 2016).
[98] R.S. KNAUER, Light suppresses melatonm secretion in humans, Science (80-. ). 210
(1980) 12.
[99] R. Yao, J. Liu, B. Li, Occupants’ adaptive responses and perception of thermal
environment in naturally conditioned university classrooms, Appl. Energy. 87 (2010)
1015–1022.

157
[100] W. O’Brien, H.B. Gunay, The contextual factors contributing to occupants’ adaptive
comfort behaviors in offices--A review and proposed modeling framework, 2014.
[101] H. Munoz, Psychological responses to light: A library center with children’s library El
Paso, Texas, (1997).
[102] R. De Dear, G.S. Brager, Developing an adaptive model of thermal comfort and
preference, (1998).
[103] D. Lindelöf, N. Morel, A field investigation of the intermediate light switching by users,
Energy Build. 38 (2006) 790–801. doi:10.1016/j.enbuild.2006.03.003.
[104] H. Gunay, W. O’Brien, I. Beausoleil-Morrison, A critical review of observation studies,
modeling, and simulation of adaptive occupant behaviors in offices, Build. Environ.
(2013). http://www.sciencedirect.com/science/article/pii/S0360132313002187 (accessed
March 31, 2016).
[105] F. Haldi, D. Robinson, On the behaviour and adaptation of office occupants, Build.
Environ. 43 (2008) 2163–2177. doi:10.1016/j.buildenv.2008.01.003.
[106] B.G. Witmer, M.J. Singer, Measuring presence in virtual environments: A presence
questionnaire, Presence. 7 (1998) 225–240.
[107] F. Hülsmann, J. Fröhlich, N. Mattar, I. Wachsmuth, Wind and warmth in virtual reality:
implementation and evaluation, in: Proc. 2014 Virtual Real. Int. Conf., ACM, 2014: p. 24.
[108] J. Dionisio, Temperature feedback in virtual environments, in: Imaging Sci. Disp.
Technol., International Society for Optics and Photonics, 1997: pp. 233–244.
[109] A. Heydarian, B. Becerik-Gerber, Use of immersive virtual environments for occupant
behaviour monitoring and data collection, J. Build. Perform. Simul. 10 (2017) 484–498.
[110] A. Heydarian, E. Pantazis, J.P. Carneiro, D. Gerber, B. Becerik-Gerber, Lights, building,
action: Impact of default lighting settings on occupant behaviour, J. Environ. Psychol. 48
(2016) 212–223.
[111] A. Heydarian, J.P. Carneiro, D. Gerber, B. Becerik-Gerber, T. Hayes, W. Wood,
Immersive virtual environments versus physical built environments: A benchmarking
study for building design and user-built environment explorations, Autom. Constr. 54

158
(2015) 116–126.
[112] S. Saeidi, T. Rizzuto, Y. Zhu, R. Kooima, Measuring the Effectiveness of an Immersive
Virtual Environment for the Modeling and Prediction of Occupant Behavior, Sustain.
Human–Building Ecosyst. (2015) 159–167.
[113] B. Deml, A. Mihalyi, G. Hannig, Development and experimental evaluation of a thermal
display, in: Proc. 2006 EuroHaptics, Citeseer, 2006: pp. 257–262.
[114] M. Bergamasco, A.A. Alessi, M. Calcara, Thermal feedback in virtual environments,
Presence Teleoperators Virtual Environ. 6 (1997) 617–629.
[115] L.A. Jones, M. Berris, The psychophysics of temperature perception and thermal-interface
design, in: Proc. 10th Symp. Haptic Interfaces Virtual Environ. Teleoperator Syst.
HAPTICS 2002, IEEE, 2002: pp. 137–142.
[116] J. Citérin, A. Pocheville, A. Kheddar, A touch rendering device in a virtual environment
with kinesthetic and thermal feedback, in: Proc. 2006 IEEE Int. Conf. Robot. Autom.
2006. ICRA 2006., IEEE, 2006: pp. 3923–3928.
[117] L.A. Jones, H.-N. Ho, Warm or cool, large or small? The challenge of thermal displays,
IEEE Trans. Haptics. 1 (2008) 53–70.
[118] G.-H. Yang, D.-S. Kwon, Effect of temperature in perceiving tactile stimulus using a
thermo-tactile display, in: 2008 Int. Conf. Control. Autom. Syst., IEEE, 2008: pp. 266–
271.
[119] M. Guiatni, A. Kheddar, Theoretical and experimental study of a heat transfer model for
thermal feedback in virtual environments, in: 2008 IEEE/RSJ Int. Conf. Intell. Robot.
Syst., IEEE, 2008: pp. 2996–3001.
[120] M. Ziat, T. Rolison, A. Shirtz, D. Wilbern, C.A. Balcer, Enhancing virtual immersion
through tactile feedback, in: Proc. Adjun. Publ. 27th Annu. ACM Symp. User Interface
Softw. Technol., ACM, 2014: pp. 65–66.
[121] R.L. Peiris, W. Peng, Z. Chen, L. Chan, K. Minamizawa, Thermovr: Exploring integrated
thermal haptic feedback with head mounted displays, in: Proc. 2017 CHI Conf. Hum.
Factors Comput. Syst., ACM, 2017: pp. 5452–5456.

159
[122] N. Ranasinghe, P. Jain, S. Karwita, D. Tolley, E.Y.-L. Do, Ambiotherm: enhancing sense
of presence in virtual reality by simulating real-world environmental conditions, in: Proc.
2017 CHI Conf. Hum. Factors Comput. Syst., ACM, 2017: pp. 1731–1742.
[123] T. Vigier, G. Moreau, D. Siret, From visual cues to climate perception in virtual urban
environments, in: 2015 IEEE Virtual Real., IEEE, 2015: pp. 305–306.
[124] H.G. Hoffman, J.N. Doctor, D.R. Patterson, G.J. Carrougher, T.A. Furness III, Virtual
reality as an adjunctive pain control during burn wound care in adolescent patients, Pain.
85 (2000) 305–309.
[125] D.R. Patterson, H.G. Hoffman, A.G. Palacios, M.J. Jensen, Analgesic effects of
posthypnotic suggestions and virtual reality distraction on thermal pain., J. Abnorm.
Psychol. 115 (2006) 834.
[126] H.G. Hoffman, D.R. Patterson, E. Seibel, M. Soltani, L. Jewett-Leahy, S.R. Sharar,
Virtual reality pain control during burn wound debridement in the hydrotank, Clin. J. Pain.
24 (2008) 299–304.
[127] J. Mott, S. Bucolo, L. Cuttle, J. Mill, M. Hilder, K. Miller, R.M. Kimble, The efficacy of
an augmented virtual reality system to alleviate pain in children undergoing burns dressing
changes: a randomised controlled trial, Burns. 34 (2008) 803–808.
[128] M. Halvey, G. Wilson, Y. Vazquez-Alvarez, S.A. Brewster, S.A. Hughes, The effect of
clothing on thermal feedback perception, in: Proc. 13th Int. Conf. Multimodal Interfaces,
ACM, 2011: pp. 217–220.
[129] J. Dionisio, Virtual hell: A trip through the flames, IEEE Comput. Graph. Appl. 17 (1997)
11–14.
[130] F. Hülsmann, N. Mattar, J. Fröhlich, I. Wachsmuth, Wind and warmth in virtual reality-
Requirements and chances, in: Proc. Work. Virtuelle Erweiterte Realität 2013, 2013.
[131] J. Fröhlich, I. Wachsmuth, Combining Multi-Sensory Stimuli in Virtual Worlds–A
Progress Report, in: Int. Conf. Virtual, Augment. Mix. Real., Springer, 2014: pp. 44–54.
[132] J. Fröhlich, I. Wachsmuth, The visual, the auditory and the haptic–a user study on
combining modalities in virtual worlds, in: Int. Conf. Virtual, Augment. Mix. Real.,

160
Springer, 2013: pp. 159–168.
[133] T. Moon, G.J. Kim, Design and evaluation of a wind display for virtual reality, in: Proc.
ACM Symp. Virtual Real. Softw. Technol., ACM, 2004: pp. 122–128.
[134] S.D. Kulkarni, M.A. Minor, M.W. Deaver, E.R. Pardyjak, J.M. Hollerbach, Design,
sensing, and control of a scaled wind tunnel for atmospheric display, IEEE/ASME Trans.
Mechatronics. 17 (2011) 635–645.
[135] S. Cardin, D. Thalmann, F. Vexo, Head mounted wind, in: Proceeding 20th Annu. Conf.
Comput. Animat. Soc. Agents, 2007: pp. 101–108.
[136] A.K. Clear, J. Morley, M. Hazas, A. Friday, O. Bates, Understanding adaptive thermal
comfort: new directions for UbiComp, in: Proc. 2013 ACM Int. Jt. Conf. Pervasive
Ubiquitous Comput., ACM, 2013: pp. 113–122.
[137] D. Yeom, J.-H. Choi, Y. Zhu, Investigation of physiological differences between
immersive virtual environment and indoor environment in a building, Indoor Built
Environ. 28 (2019) 46–62.
[138] J. Van Hoof, Abstract, Indoor Air. 18 (2008) 182–201.
[139] H. Hensel, Neural processes in thermoregulation., Physiol. Rev. 53 (1973) 948–1017.
[140] M. Cabanac, Thermoregulatory behavior, J. Physiol. (Paris). 75 (1979) 115–178.
[141] J.D. Hardy, Thermal comfort: skin temperature and physiological thermoregulation,
Physiol. Behav. Temp. Regul. (1970) 856–873.
[142] R.A. Bakar, N. Jusoh, A.R. Ismail, T.Z.S. Ali, Effect on human metabolic rate of skin
temperature in an office occupant, in: MATEC Web Conf., EDP Sciences, 2017: p. 1070.
[143] D. Wang, J. Jiang, Y. Liu, Y. Wang, Y. Xu, J. Liu, Student responses to classroom
thermal environments in rural primary and secondary schools in winter, Build. Environ.
115 (2017) 104–117.
[144] K.C. Parsons, Environmental ergonomics: a review of principles, methods and models,
Appl. Ergon. 31 (2000) 581–594.
[145] I.S. Walker, A.K. Meier, Residential Thermostats : Comfort Controls in California Homes

161
Iain S . Walker and Alan K . Meier Lawrence Berkeley National Laboratory Berkeley ,
California March 2008 LBNL-938E, (2008).
[146] M. Veselý, P. Molenaar, M. Vos, R. Li, W. Zeiler, Personalized heating–Comparison of
heaters and control modes, Build. Environ. 112 (2017) 223–232.
[147] T. Parkinson, R. de Dear, Thermal pleasure in built environments: spatial alliesthesia from
air movement, Build. Res. Inf. 45 (2017) 320–335.
[148] T. Bedford, The Warmth Factor in Comfort at Work. A Physiological Study of Heating
and Ventilation., Warmth Factor Comf. Work. A Physiol. Study Heat. Vent. (1936).
[149] Z.J. Schlader, H.D. Prange, T.D. Mickleborough, J.M. Stager, Characteristics of the
control of human thermoregulatory behavior, Physiol. Behav. 98 (2009) 557–562.
[150] B. Koelblen, A. Psikuta, A. Bogdan, S. Annaheim, R.M. Rossi, Thermal sensation
models: a systematic comparison, Indoor Air. 27 (2017) 680–689.
[151] M. Schweiker, X. Fuchs, S. Becker, M. Shukuya, M. Dovjak, M. Hawighorst, J. Kolarik,
Challenging the assumptions for thermal sensation scales, Build. Res. Inf. 45 (2017) 572–
589.
[152] L.R. Glicksman, S. Taub, Thermal and behavioral modeling of occupant-controlled
heating, ventilating and air conditioning systems, Energy Build. 25 (1997) 243–249.
[153] A.C. Boerstra, T.C. Beuker, Impact of perceived personal control over indoor climate on
health and comfort in Dutch offices, Indoor Air. (2011).
[154] R.V. Andersen, B.W. Olesen, J. Toftum, Modelling occupants’ heating set-point
prefferences, in: Build. Simul. 2011 12th Conf. Int. Build. Perform. Simul. Assoc., 2011.
[155] J.F. Nicol, M.A. Humphreys, Adaptive thermal comfort and sustainable thermal standards
for buildings, Energy Build. 34 (2002) 563–572. doi:10.1016/S0378-7788(02)00006-3.
[156] S. Shahzad, J. Brennan, D. Theodossopoulos, B. Hughes, J.K. Calautit, A study of the
impact of individual thermal control on user comfort in the workplace: Norwegian cellular
vs. British open plan offices, Archit. Sci. Rev. 60 (2017) 49–61.
[157] A.K. Melikov, B. Krejcirikova, J. Kaczmarczyk, M. Duszyk, T. Sakoi, Human response to
local convective and radiant cooling in a warm environment, Hvac&r Res. 19 (2013)

162
1023–1032.
[158] A.K. Melikov, G.L. Knudsen, Human response to an individually controlled
microenvironment, HVAC&R Res. 13 (2007) 645–660.
[159] F. Kalmár, Innovative method and equipment for personalized ventilation, Indoor Air. 25
(2015) 297–306.
[160] S. Schiavon, B. Yang, Y. Donner, V. Chang, W.W. Nazaroff, Thermal comfort, perceived
air quality, and cognitive performance when personally controlled air movement is used
by tropically acclimatized persons, Indoor Air. 27 (2017) 690–702.
[161] M. He, N. Li, Y. He, D. He, C. Song, The influence of personally controlled desk fan on
comfort and energy consumption in hot and humid environments, Build. Environ. 123
(2017) 378–389.
[162] C. Chun, A. Kwok, T. Mitamura, N. Miwa, A. Tamura, Thermal diary: Connecting
temperature history to indoor comfort, Build. Environ. 43 (2008) 877–885.
[163] J. Yu, G. Cao, W. Cui, Q. Ouyang, Y. Zhu, People who live in a cold climate: thermal
adaptation differences based on availability of heating, Indoor Air. 23 (2013) 303–310.
[164] P.O. Fanger, Thermal comfort. Analysis and applications in environmental engineering.,
Therm. Comf. Anal. Appl. Environ. Eng. (1970).
[165] H. Hu, R. Wang, C. Zhao, H. Luo, L. Ding, Y. Qiu, Experimental study on thermal
comfort of indoor environment, in: Adv. Ergon. Model. Usability Spec. Popul., Springer,
2017: pp. 391–401.
[166] L.T. Wong, K.W. Mui, C.T. Cheung, Bayesian thermal comfort model, Build. Environ. 82
(2014) 171–179.
[167] H. Zhang, C. Huizenga, E. Arens, T. Yu, Considering individual physiological differences
in a human thermal model, J. Therm. Biol. 26 (2001) 401–408. doi:10.1016/S0306-
4565(01)00051-1.
[168] G. Havenith, I. Holmér, K. Parsons, Personal factors in thermal comfort assessment:
Clothing properties and metabolic heat production, Energy Build. 34 (2002) 581–591.
doi:10.1016/S0378-7788(02)00008-7.

163
[169] M. De Carli, B.W. Olesen, A. Zarrella, R. Zecchin, People’s clothing behaviour according
to external weather and indoor environment, Build. Environ. 42 (2007) 3965–3973.
[170] K.N. Nkurikiyeyezu, Y. Suzuki, G.F. Lopez, Heart rate variability as a predictive
biomarker of thermal comfort, J. Ambient Intell. Humaniz. Comput. 9 (2018) 1465–1477.
[171] A. Ghahramani, C. Tang, B. Becerik-Gerber, An online learning approach for quantifying
personalized thermal comfort via adaptive stochastic modeling, Build. Environ. 92 (2015)
86–96.
[172] K. Van Den Wymelenberg, Patterns of occupant interaction with window blinds: A
literature review, Energy Build. 51 (2012) 165–176. doi:10.1016/j.enbuild.2012.05.008.
[173] M. Rea, Window blind occlusion: a pilot study, Build. Environ. (1984). http://nparc.cisti-
icist.nrc-cnrc.gc.ca/npsi/ctrl?action=rtdoc&an=20358948 (accessed April 2, 2016).
[174] M. Foster, T. Oreszczyn, Occupant control of passive systems: The use of Venetian
blinds, Build. Environ. 36 (2001) 149–155. doi:10.1016/S0360-1323(99)00074-8.
[175] Y.-C. Chan, A. Tzempelikos, Efficient venetian blind control strategies considering
daylight utilization and glare protection, Sol. Energy. 98 (2013) 241–254.
[176] K.G. Van Den Wymelenberg, Visual Comfort, Discomfort Glare, and Occupant
Fenestration Control: Developing a Research Agenda, Leukos. 10 (2014) 207–221.
doi:10.1080/15502724.2014.939004.
[177] J. Day, J. Theodorson, Understanding controls, behaviors and satisfaction in the daylit
perimeter office: A daylight design case study, J. Inter. (2012).
http://onlinelibrary.wiley.com/doi/10.1111/j.1939-1668.2011.01068.x/full (accessed April
2, 2016).
[178] A. Mahdavi, Patterns and implications of user control actions in buildings, Indoor Built
Environ. 18 (2009) 440–446. doi:10.1177/1420326X09344277.
[179] A. Mahdavi, A. Mohammadi, E. Kabir, L. Lambeva, Occupants’ operation of lighting and
shading systems in office buildings, J. Build. Perform. Simul. 1 (2008) 57–65.
doi:10.1080/19401490801906502.
[180] V. Inkarojrit, Multivariate prdictive window blind control models for intelligent building

164
facade systems, Proc. Build. Simul. (2007) 787–794.
http://www.ibpsa.org/proceedings/BS2007/p519_final.pdf.
[181] 井上, The development of an optimal control system for window shading devices based
on investigation in office buildings, ASHRAE Trans. 94 (1988).
[182] C.F. Reinhart, K. Voss, Monitoring manual control of electric lighting and blinds, Light.
Res. Technol. 35 (2003) 243–258.
[183] B. Meerbeek, M. te Kulve, T. Gritti, M. Aarts, E. van Loenen, E. Aarts, Building
automation and perceived control: A field study on motorized exterior blinds in Dutch
offices, Build. Environ. 79 (2014) 66–77. doi:10.1016/j.buildenv.2014.04.023.
[184] H.B. Gunay, W. O’Brien, I. Beausoleil-Morrison, B. Huchuk, On adaptive occupant-
learning window blind and lighting controls, Build. Res. Inf. (2014) 1–18.
[185] Y. Zhang, P. Barrett, Factors influencing occupants’ blind-control behaviour in a naturally
ventilated office building, Build. Environ. 54 (2012) 137–147.
[186] M.V. Bavaresco, E. Ghisi, Influence of user interaction with internal blinds on the energy
efficiency of office buildings, Energy Build. 166 (2018) 538–549.
doi:10.1016/j.enbuild.2018.02.011.
[187] C.F. Reinhart, Daylight availability and manual lighting control in office buildings:
simulation studies and analysis of measurement, Fraunhofer-IRB-Verlag, 2001.
[188] Y. Sutter, D. Dumortier, M. Fontoynont, The use of shading systems in VDU task offices:
A pilot study, Energy Build. 38 (2006) 780–789.
[189] L. Sanati, M. Utzinger, The effect of window shading design on occupant use of blinds
and electric lighting, Build. Environ. 64 (2013) 67–76.
[190] A. Nezamdoost, K. Van Den Wymelenberg, Blindswitch 2017: Proposing a new manual
blind control algorithm for daylight and energy simulation, in: IES Annu. Conf.
Proceedings. Portland, OR, 2017.
[191] E.S. Lee, D.L. DiBartolomeo, S.E. Selkowitz, Thermal and daylighting performance of an
automated venetian blind and lighting system in a full-scale private office, Energy Build.
29 (1998) 47–63. doi:10.1016/S0378-7788(98)00035-8.

165
[192] I. Bennet, W. O’Brien, H.B. Gunay, Effect of window blind use in residential buildings:
Observation and simulation study, Proc. ESIM2014, Ottawa, Canada. 89 (2014).
[193] S.A.M. Offermans, H.A. van Essen, J.H. Eggen, User interaction with everyday lighting
systems, Pers. Ubiquitous Comput. 18 (2014) 2035–2055. doi:10.1007/s00779-014-0759-
2.
[194] S. Escuyer, M. Fontoynont, Lighting controls: a field study of office workers’ reactions,
Light. Res. Technol. 33 (2001) 77–94.
[195] F. Naspi, M. Arnesano, L. Zampetti, F. Stazi, G.M. Revel, M. D’Orazio, Experimental
study on occupants’ interaction with windows and lights in Mediterranean offices during
the non-heating season, Build. Environ. 127 (2018) 221–238.
doi:10.1016/j.buildenv.2017.11.009.
[196] P.C. da Silvaa, V. Leala, M. Andersenb, OCCUPANTS INTERACTION WITH
ELECTRIC LIGHTING AND SHADING SYSTEMS IN REAL SINGLE-OCCUPIED
OFFICES: RESULTS FROM A MONITORING CAMPAIGN, (n.d.).
[197] E. Vine, E. Lee, R. Clear, D. DiBartolomeo, S. Selkowitz, Office worker response to an
automated venetian blind and electric lighting system: a pilot study, Energy Build. 28
(1998) 205–218.
[198] V. Fabi, V. Camisassi, F. Causone, S.P. Corgnati, R. Andersen, Light switch behaviour :
occupant behaviour stochastic models in office buildings, 8th Wind. Conf. Count. Cost
Comf. a Chang. World. 8 (2014) 10–13.
[199] J. Langevin, J. Wen, P.L. Gurian, Quantifying the human–building interaction:
Considering the active, adaptive occupant in building performance simulation, Energy
Build. 117 (2016) 372–386.
[200] S. Wei, R. Jones, P. De Wilde, Driving factors for occupant-controlled space heating in
residential buildings, Energy Build. 70 (2014) 36–44.
[201] A. Meier, C. Aragon, B. Hurwitz, D. Mujumdar, T. Peffer, D. Perry, M. Pritoni, How
people actually use thermostats, Lawrence Berkeley National Lab.(LBNL), Berkeley, CA
(United States), 2010.

166
[202] B. Urban, D. Elliott, O. Sachs, Towards Better Modeling of Residential Thermostats,
Proc. SimBuild. 5 (2012) 56–62.
[203] X. Chen, J. Jang, D.M. Auslander, T. Peffer, E.A. Arens, Demand response-enabled
residential thermostat controls., (2008).
[204] M. Veselý, W. Zeiler, Personalized conditioning and its impact on thermal comfort and
energy performance - A review, Renew. Sustain. Energy Rev. 34 (2014) 401–408.
doi:10.1016/j.rser.2014.03.024.
[205] M. Veselý, Y. Zhao, M. Vos, W. Zeiler, Process Control of Personalized Heating, Proc.
9th Wind. Conf. (2016) 7–10.
[206] A. Bogdan, M. Chludzinska, A. Bogdan, Assessment of Thermal Comfort Using
Personalized Ventilation Assessment of Thermal Comfort Using Personalized Ventilation,
9669 (2016). doi:10.1080/10789669.2010.10390919.
[207] M. Schweiker, S. Brasche, W. Bischof, M. Hawighorst, K. Voss, A. Wagner,
Development and validation of a methodology to challenge the adaptive comfort model,
Build. Environ. 49 (2012) 336–347.
[208] M. Schweiker, S. Brasche, W. Bischof, M. Hawighorst, A. Wagner, Explaining the
individual processes leading to adaptive comfort: Exploring physiological, behavioural
and psychological reactions to thermal stimuli, J. Build. Phys. 36 (2013) 438–463.
[209] S. Tanabe, M. Haneda, N. Nishihara, Workplace productivity and individual thermal
satisfaction, Build. Environ. 91 (2015) 42–50.
[210] C.-Y. Hsu, M. Donn, Commercial building façade design: the relationship between early
design lessons and detailed design lessons, (2009).
[211] D. Lindelöf, N. Morel, Bayesian estimation of visual discomfort, Build. Res. Inf. 36
(2008) 83–96.
[212] A.D. Galasiu, J.A. Veitch, Occupant preferences and satisfaction with the luminous
environment and control systems in daylit offices: a literature review, Energy Build. 38
(2006) 728–742.
[213] M.-C. Dubois, Shading devices and daylight quality: an evaluation based on simple

167
performance indicators, Light. Res. Technol. 35 (2003) 61–74.
[214] A. Borisuit, F. Linhart, J.-L. Scartezzini, M. Münch, Effects of realistic office daylighting
and electric lighting conditions on visual comfort, alertness and mood, Light. Res.
Technol. 47 (2015) 192–209.
[215] L. Xia, Y. Tu, L. Liu, Y. Wang, S. Peng, M. Knoop, I. Heynderickx, A study on overhead
glare in office lighting conditions, J. Soc. Inf. Disp. 19 (2011) 888–898.
[216] J.A. Clarke, M. Janak, P. Ruyssevelt, Assessing the overall performance of advanced
glazing systems, Sol. Energy. 63 (1998) 231–241.
[217] Y. Bian, Y. Ma, Subjective survey & simulation analysis of time-based visual comfort in
daylit spaces, Build. Environ. 131 (2018) 63–73.
[218] G. Parise, L. Martirano, Combined electric light and daylight systems ecodesign, IEEE
Trans. Ind. Appl. 49 (2013) 1062–1070.
[219] H. Shen, A. Tzempelikos, Daylighting and energy analysis of private offices with
automated interior roller shades, Sol. Energy. 86 (2012) 681–704.
doi:10.1016/j.solener.2011.11.016.
[220] G.R. Newsham, M.B.C. Aries, S. Mancini, G. Faye, Individual control of electric lighting
in a daylit space, Light. Res. Technol. 40 (2008) 25–41.
[221] H.B. Gunay, W. O’Brien, I. Beausoleil-Morrison, S. Gilani, Development and
implementation of an adaptive lighting and blinds control algorithm, Build. Environ. 113
(2017) 185–199.
[222] N. Kunwar, K.S. Cetin, U. Passe, Dynamic Shading in Buildings: a Review of Testing
Methods and Recent Research Findings, Curr. Sustain. Energy Reports. 5 (2018) 93–100.
[223] V. Garcia-Hansen, A.C. Allan, G. Isoardi, A. Wagdy, S.S. Smith, Evaluating visual
comfort in open-plan offices: Exploration of simple methods for evaluation and
prediction, (2017).
[224] S. Hoffmann, E.S. Lee, A. McNeil, L. Fernandes, D. Vidanovic, A. Thanachareonkit,
Balancing daylight, glare, and energy-efficiency goals: An evaluation of exterior coplanar
shading systems using complex fenestration modeling tools, Energy Build. 112 (2016)

168
279–298.
[225] C.-M. Lai, S. Hokoi, Solar façades: A review, Build. Environ. 91 (2015) 152–165.
[226] M.M.S. Ahmed, A.K. Abel-Rahman, A.H.H. Ali, Development of intelligent façade based
on outdoor environment and indoor thermal comfort, Procedia Technol. 19 (2015) 742–
749.
[227] L. Vanhoutteghem, G.C.J. Skarning, C.A. Hviid, S. Svendsen, Impact of façade window
design on energy, daylighting and thermal comfort in nearly zero-energy houses, Energy
Build. 102 (2015) 149–156.
[228] M. Kragh, M. Colombari, M. Zobec, Advanced Façades and HVAC Systems: Preliminary
results of full-scale monitoring, Energy Effic. Heal. Build. Sustain. Cities. (2002).
[229] Y. Huang, J. Niu, T. Chung, Comprehensive analysis on thermal and daylighting
performance of glazing and shading designs on office building envelope in cooling-
dominant climates, Appl. Energy. 134 (2014) 215–228.
[230] L. Karlsen, P. Heiselberg, I. Bryn, H. Johra, Solar shading control strategy for office
buildings in cold climate, Energy Build. 118 (2016) 316–328.
[231] F. Biocca, The cyborg’s dilemma: Progressive embodiment in virtual environments, J.
Comput. Commun. 3 (1997) JCMC324.
[232] M. Slater, A. Steed, A virtual presence counter, Presence Teleoperators Virtual Environ. 9
(2000) 413–434.
[233] E.K. Park, S.S. Sundar, Can synchronicity and visual modality enhance social presence in
mobile messaging?, Comput. Human Behav. 45 (2015) 121–128.
[234] K.L. Nowak, F. Biocca, The effect of the agency and anthropomorphism on users’ sense
of telepresence, copresence, and social presence in virtual environments, Presence
Teleoperators Virtual Environ. 12 (2003) 481–494.
[235] M. Slater, Real people meeting virtually real people—a review of some experiments in
shared virtual environments, BT Technol. J. 17 (1999) 120–127.
[236] R. Schroeder, A. Steed, A.-S. Axelsson, I. Heldal, Å. Abelin, J. Wideström, A. Nilsson,
M. Slater, Collaborating in networked immersive spaces: as good as being there together?,

169
Comput. Graph. 25 (2001) 781–788.
[237] M. Lombard, T. Ditton, At the heart of it all: The concept of presence, J. Comput.
Commun. 3 (1997) JCMC321.
[238] M. Garau, M. Slater, D.-P. Pertaub, S. Razzaque, The responses of people to virtual
humans in an immersive virtual environment, Presence Teleoperators Virtual Environ. 14
(2005) 104–116.
[239] S. Pellegrini, A. Ess, K. Schindler, L. Van Gool, You’ll never walk alone: Modeling social
behavior for multi-target tracking, in: Comput. Vision, 2009 IEEE 12th Int. Conf., IEEE,
2009: pp. 261–268.
[240] D. Helbing, P. Molnar, Social force model for pedestrian dynamics, Phys. Rev. E. 51
(1995) 4282.
[241] S. Ali, M. Shah, Floor fields for tracking in high density crowd scenes, in: Eur. Conf.
Comput. Vis., Springer, 2008: pp. 1–14.
[242] F. Haldi, D. Robinson, Adaptive actions on shading devices in response to local visual
stimuli, J. Build. Perform. Simul. 3 (2010) 135–153. doi:10.1080/19401490903580759.
[243] J.K. Day, Occupant Behaviors and Energy Use: Creating High-Performance People for
High-Performance Buildings, in: Sustain. Human–Building Ecosyst., 2015: pp. 18–26.
[244] J.F. Nicol, M. Humphreys, Understanding the adaptive approach to thermal comfort,
ASHRAE Trans. 104 (1998) 991–1004.
[245] J.A. Clarke, I.A. Macdonald, J.F. Nicol, Predicting adaptive responses-simulating
occupied environments, (2006).
[246] H.B. Rijal, P. Tuohy, M.A. Humphreys, J.F. Nicol, A. Samuel, An algorithm to represent
occupant use of windows and fans including situation-specific motivations and
constraints, Build. Simul. 4 (2011) 117–134. doi:10.1007/s12273-011-0037-4.
[247] M. Konstantoglou, A. Tsangrassoulis, Dynamic operation of daylighting and shading
systems: A literature review, Renew. Sustain. Energy Rev. 60 (2016) 268–283.
[248] F. Stazi, F. Naspi, M. D’Orazio, A literature review on driving factors and contextual
events influencing occupants’ behaviours in buildings, Build. Environ. 118 (2017) 40–66.

170
[249] G.R. Newsham, Manual control of window blinds and electric lighting: implications for
comfort and energy consumption, Indoor Environ. 3 (1994) 135–144.
[250] S. Olbina, Split controlled blinds as a thermal and daylighting environmental control
system, in: Proc. 3rd CIB Int. Conf. Smart Sustain. Built Environ. (SASBE2009), Delft,
Netherlands, 2009.
[251] G.S.B. I, R.J. De Dear, Thermal adaptation in the built environment : a literature review,
Energy Build. 27 (1998) 83–96. doi:10.1016/S0378-7788(97)00053-4.
[252] G. Jendritzky, R. de Dear, Adaptation and thermal environment, in: Biometeorol. Adapt.
to Clim. Var. Chang., Springer, 2009: pp. 9–32.
[253] J.H. Choi, A. Aziz, V. Loftness, Investigation on the impacts of different genders and ages
on satisfaction with thermal environments in office buildings, Build. Environ. 45 (2010)
1529–1535. doi:10.1016/j.buildenv.2010.01.004.
[254] S. Karjalainen, Gender differences in thermal comfort and use of thermostats in everyday
thermal environments, Build. Environ. 42 (2007) 1594–1603.
doi:10.1016/j.buildenv.2006.01.009.
[255] H. Zhang, E. Arens, C. Huizenga, T. Han, Thermal sensation and comfort models for non-
uniform and transient environments: Part I: Local sensation of individual body parts,
Build. Environ. 45 (2010) 380–388.
[256] Z. Mo, A. Mahdavi, An agent-based simulation-assisted approach to bi-lateral building
systems control, na, 2003.
[257] E.H. Mathews, C.P. Botha, D.C. Arndt, A. Malan, HVAC control strategies to enhance
comfort and minimise energy usage, Energy Build. 33 (2001) 853–863.
[258] E.C.M. van Oeffelen, K. van Zundert, P. Jacobs, Persoonlijke verwarming in
kantoorgebouwen, TVVL Mag. Januari, 1, 39, 6-11. (2010).
[259] S. Ari, I.A. Cosden, H.E. Khalifa, J.F. Dannenhoffer, P. Wilcoxen, C. Isik, Individual
thermal comfort and energy optimization, Proc. Clima Wellbeing Indoors. (2007).
[260] A. Roetzel, A. Tsangrassoulis, U. Dietrich, S. Busching, Context dependency of comfort
and energy performance in mixed-mode offices, J. Build. Perform. Simul. 4 (2011) 303–

171
322.
[261] K. Sun, T. Hong, A simulation approach to estimate energy savings potential of occupant
behavior measures, Energy Build. 136 (2017) 43–62.
[262] Y. Zhang, X. Bai, F.P. Mills, J.C. V Pezzey, Rethinking the role of occupant behavior in
building energy performance: A review, Energy Build. 172 (2018) 279–294.
[263] J. Grabe, C. Gonzalez, Human decision making in energy-relevant interaction with
buildings, in: Cent. Eur. Symp. Build. Phys. (CESBP 2016), 2016: pp. 14–16.
[264] R.J. De Dear, G.S. Brager, Thermal comfort in naturally ventilated buildings: revisions to
ASHRAE Standard 55, Energy Build. 34 (2002) 549–561.
[265] R.J. de Dear, T. Akimoto, E.A. Arens, G. Brager, C. Candido, K.W.D. Cheong, B. Li, N.
Nishihara, S.C. Sekhar, S. Tanabe, Progress in thermal comfort research over the last
twenty years, Indoor Air. 23 (2013) 442–461.
[266] D.A. McIntyre, Chamber studies—reductio ad absurdum?, Energy Build. 5 (1982) 89–96.
[267] C. Benton, F. Bauman, M. Fountain, A field measurement system for the study of thermal
comfort, (1990).
[268] R. de Dear, M. Fountain, Field experiments on occupant comfort and office thermal
environments in a hot-humid climate, (1994).
[269] K. Cena, R. de Dear, Thermal comfort and behavioural strategies in office buildings
located in a hot-arid climate, J. Therm. Biol. 26 (2001) 409–414.
[270] M. Slater, M. Usoh, Representations systems, perceptual position, and presence in
immersive virtual environments, Presence Teleoperators Virtual Environ. 2 (1993) 221–
233.
[271] M.F. Kuntze, R. Stoermer, R. Mager, A. Roessler, F. Mueller-Spahn, A.H. Bullinger,
Immersive virtual environments in cue exposure, Cyberpsychology Behav. 4 (2001) 497–
501.
[272] K.F. Hew, W.S. Cheung, Use of three‐dimensional (3‐D) immersive virtual worlds in K‐
12 and higher education settings: A review of the research, Br. J. Educ. Technol. 41
(2010) 33–55.

172
[273] J. Goulding, W. Nadim, P. Petridis, M. Alshawi, Construction industry offsite production:
A virtual reality interactive training environment prototype, Adv. Eng. Informatics. 26
(2012) 103–116.
[274] H.-J. Bullinger, W. Bauer, G. Wenzel, R. Blach, Towards user centred design (UCD) in
architecture based on immersive virtual environments, Comput. Ind. 61 (2010) 372–379.
[275] S. Woksepp, T. Olofsson, Credibility and applicability of virtual reality models in design
and construction, Adv. Eng. Informatics. 22 (2008) 520–528.
[276] L. Sastry, D.R.S. Boyd, Virtual environments for engineering applications, Virtual Real. 3
(1998) 235–244.
[277] Y. Yang, B. Li, H. Liu, M. Tan, R. Yao, A study of adaptive thermal comfort in a well-
controlled climate chamber, Appl. Therm. Eng. 76 (2015) 283–291.
[278] Q. Zhao, Z. Cheng, F. Wang, Z. Chen, Y. Jiang, Z. Zhong, Experimental assessment of a
satisfaction based thermal comfort control for a group of occupants, in: 2015 IEEE Int.
Conf. Autom. Sci. Eng., IEEE, 2015: pp. 15–20.
[279] Q. Jin, L. Duanmu, Experimental study of thermal sensation and physiological response
during step changes in non-uniform indoor environment, Sci. Technol. Built Environ. 22
(2016) 237–247.
[280] A. Jelfs, D. Whitelock, Presence and the role of activity theory in understanding: How
students learn in virtual learning environments, in: Int. Conf. Cogn. Technol., Springer,
2001: pp. 123–129.
[281] B.A. Nardi, Context and consciousness: activity theory and human-computer interaction,
mit Press, 1996.
[282] G.R. Newsham, Clothing as a thermal comfort moderator and the effect on energy
consumption, Energy Build. 26 (1997) 283–291. doi:10.1016/S0378-7788(97)00009-1.
[283] M. Slater, M. Usoh, A. Steed, Depth of presence in virtual environments, Presence
Teleoperators Virtual Environ. 3 (1994) 130–144.
[284] G.M. Revel, M. Arnesano, Measuring overall thermal comfort to balance energy use in
sports facilities, Measurement. 55 (2014) 382–393.

173
[285] N. Walikewitz, B. Jänicke, M. Langner, F. Meier, W. Endlicher, The difference between
the mean radiant temperature and the air temperature within indoor environments: A case
study during summer conditions, Build. Environ. 84 (2015) 151–161.
[286] P. Antoniadou, E. Kyriaki, A. Manoloudis, A.M. Papadopoulos, Evaluation of Thermal
Sensation in Office Buildings: A case study in the Mediterranean, Procedia Environ. Sci.
38 (2017) 28–35.
[287] A. Matzarakis, B. Amelung, Physiological equivalent temperature as indicator for impacts
of climate change on thermal comfort of humans, in: Seas. Forecast. Clim. Chang. Hum.
Heal., Springer, 2008: pp. 161–172.
[288] O. Kaynakli, M. Kilic, Investigation of indoor thermal comfort under transient conditions,
Build. Environ. 40 (2005) 165–174.
[289] M. Langner, K. Scherber, W.R. Endlicher, Indoor heat stress: An assessment of human
bioclimate using the UTCI in different buildings in Berlin, DIE ERDE–Journal Geogr.
Soc. Berlin. 144 (2014) 260–273.
[290] K.M. Stanney, R.R. Mourant, R.S. Kennedy, Human factors issues in virtual
environments: A review of the literature, Presence. 7 (1998) 327–351.
[291] G. Ozcelik, B. Becerik-Gerber, Benchmarking thermoception in virtual environments to
physical environments for understanding human-building interactions, Adv. Eng.
Informatics. 36 (2018) 254–263. doi:10.1016/j.aei.2018.04.008.
[292] J.A. Veitch, How and why to assess workplace design: facilities management supports
human resources, Organ. Dyn. (2018).
[293] A.R. Pollard, R. O’Driscoll, D.N. Pinder, The impact of solar radiation on the air
temperature within a residential building, BRANZ, 2001.
[294] G. Ozcelik, B. Becerik-Gerber, R. Chugh, Interactive Visualizations, (2018).
doi:10.5281/ZENODO.2483288.
[295] S.J. Roskam, Brightness Perception and the Effect of Synthetic Glare in Virtual Lighting
Applications, (2015).
[296] J. Kolkmeier, J. Vroon, D. Heylen, Interacting with virtual agents in shared space: Single

174
and joint effects of gaze and proxemics, in: Int. Conf. Intell. Virtual Agents, Springer,
2016: pp. 1–14.
[297] G.S. Brager, R. De Dear, Climate, comfort, & natural ventilation: a new adaptive comfort
standard for ASHRAE standard 55, (2001).
[298] C. Wang, D. Yan, H. Sun, Y. Jiang, A generalized probabilistic formula relating occupant
behavior to environmental conditions, Build. Environ. 95 (2016) 53–62.
doi:10.1016/j.buildenv.2015.09.004.
[299] D.R.G. Hunt, The use of artificial lighting in relation to daylight levels and occupancy,
Build. Environ. 14 (1979) 21–33.
[300] J.F. Nicol, M.A. Humphreys, A Stochastic Approach to Thermal Comfort-Occupant
Behavior and Energy Use in Buildings/DISCUSSION, ASHRAE Trans. 110 (2004) 554.
[301] F. Haldi, D. Robinson, On the unification of thermal perception and adaptive actions,
Build. Environ. 45 (2010) 2440–2457. doi:10.1016/j.buildenv.2010.05.010.
[302] V. Fabi, R.V. Andersen, S. Corgnati, B.W. Olesen, Occupants’ window opening
behaviour: A literature review of factors influencing occupant behaviour and models,
Build. Environ. 58 (2012) 188–198. doi:10.1016/j.buildenv.2012.07.009.
[303] G. Ozcelik, B. Becerik-Gerber, Reasons of Human-Building Interactions (HBIs), (2019).
doi:10.5281/ZENODO.3245063.

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Enabling human-building communication to promote pro-environmental behavior in office buildings

PDF

Understanding human-building-emergency interactions in the built environment

PDF

Intelligent adaptive automation: activity-driven and user-centered building automation

PDF

Building occupancy modeling and occupancy-loads relationships for building heating/cooling energy efficiency

PDF

User-centric smart sensing for non-intrusive electricity consumption disaggregation in buildings

PDF

Towards health-conscious spaces: building for human well-being and performance

PDF

Modeling the spatio-temporal variability of solar radiation on buildings: a case study of Lewis Hall

PDF

In-situ quality assessment of scan data for as-built models using building-specific geometric features

PDF

Behavioral form finding using multi-agent systems: a computational methodology for combining generative design with environmental and structural analysis in architectural design

PDF

Bounded technological rationality: the intersection between artificial intelligence, cognition, and environment and its effects on decision-making

PDF

Real-time simulation-based feedback on carbon impacts for user-engaged temperature management

PDF

Essays on improving human interactions with humans, algorithms, and technologies for better healthcare outcomes

PDF

Decoding information about human-agent negotiations from brain patterns

PDF

A parametric study of the thermal performance of green roofs in different climates through energy modeling

PDF

Enhancing thermal comfort: air temperature control based on human facial skin temperature

PDF

Distributed adaptive control with application to heating, ventilation and air-conditioning systems

Asset Metadata

Core Title Understanding human-building interactions through perceptual decision-making processes

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Civil Engineering

Degree Conferral Date 2019-08

Publication Date 08/12/2021

Defense Date 05/21/2019

Tag adaptive and responsive environments,built environments,cognitive environments,commercial buildings,energy consumption,energy consumption behavior,human subject research,human-building interaction,human-computer interactions,immersive virtual environments,multimodal discomfort,multimodal perceptions,multi-occupancy offices,OAI-PMH Harvest,occupant behavior,occupant comfort,occupant decision-making,occupant interactions,occupant response,occupant satisfaction,perceptions,perceptual decision-making,reasons of interactions,single occupancy offices,smart environments,sustainability,thermoception,user experience,virtual agents,virtual reality

Language English

Advisor Becerik-Gerber, Burcin (committee chair), Lucas, Gale (committee member), Soibelman, Lucio (committee member)

Creator Email gokceozcelik92@gmail.com,gozcelik@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-214304

Unique identifier UC11663034

Legacy Identifier etd-OzcelikGok-7769

Dmrecord 214304

Document Type Dissertation

Rights Ozcelik, Gokce

Source University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu