Facebook habits: rewards, cues, and automaticity

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Request accessible transcript

Transcript (if available)

Content

Facebook Habits:
Rewards, Cues, and Automaticity

by
Ian A. Anderson

A Thesis Presented to the
FACULTY OF THE USC DORNSIFE
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF ARTS
(PSYCHOLOGY)

AUGUST 2021

Copyright 2021 Ian Anderson

ii
Acknowledgements
The authors thank Alan Mislove for graciously sharing the data for Study 2 (September 9,
2020), Steve Read and Norbert Schwarz for their comments on an earlier draft of this work, the
USC habit lab for advice on study design and analysis, and Hannah Upshaw, Julie Cho, Elyse
Manzo, Yashar Haghighiara, Alyssa Manley, Rachel Cai, Jeremy Barajas, Manasi Iyer, Denver
Yu, and Sarah Lee for their assistance in collecting data for Study 1. Correspondence should be
addressed to Ian Anderson, University of Southern California, Department of Psychology, 3551
Trousdale Pkwy, Los Angeles, CA. 90089. Email: ianaxelanderson@gmail.com

iii
Table of Contents
Acknowledgements .................................................................................................................. ii
List of Tables ............................................................................................................................ iv
List of Figures ............................................................................................................................ v
Abstract ..................................................................................................................................... vi
Preface: Statement of Relevance ............................................................................................ vii
Facebook Habits: Rewards, Cues, and Automaticity ................................................................ 1
Study 1: Rewards Motivate Initial/Occasional Facebook Use .................................................. 3
Method ................................................................................................................................... 3
Results .................................................................................................................................... 6
Discussion ............................................................................................................................ 12
Study 2: Cues Disrupt Habitual Users of Facebook ................................................................ 13
Method ................................................................................................................................. 13
Results .................................................................................................................................. 15
Discussion ............................................................................................................................ 19
Study 3: Predictions about Posting Given Rewards and Platform Changes ............................ 20
Method ................................................................................................................................. 20
Results .................................................................................................................................. 22
Discussion ............................................................................................................................ 29
General Discussion .................................................................................................................. 29
References ............................................................................................................................... 31

iv
List of Tables
Table 1: Means, Standard Deviations, and Correlations from Study 1…………….…………..6
Table 2: Study 1 Multilevel Model Predicting Latency Between Posts From Past Posting
Frequency and Social Rewards……………………………………………………….………..8
Table 3: Study 1 Multilevel Model Predicting Latency Between Posts From SRHI and Social
Rewards……………………………………………………………………………….……….10
Table 4: Study 2 Means, Standard Deviations, and Correlation…………………….…………15
Table 5: Results of Multilevel Regression Model Predicting Latency Between Facebook Posts As
A Function of A Facebook Platform Change and Past Posting Frequency: Study 2………….16
Table 6: Study 3 Means, Standard Deviations, and Correlations…………………………...…22

v
List of Figures
Figure 1: Plot of the Interaction in Study 1 Between Past Weekly Posts and Social Rewards
Predicting Latency Between Subsequent Posts…………………………………………………9
Figure 2: Plot of the Interaction in Study 1 Between Self-Report Habit Index (SRHI) and Social
Rewards Predicting Latency Between Subsequent Posts………………………………………11
Figure 3: Plot of Latency Between Facebook Posts As A Function of the Timing of A Platform
Change and Past Posting Frequency; Study 2………………………………………………….18
Figure 4: Predictions of Latency to Post Given Social Rewards As A Function of Prior Post
Frequency in Study 3…………………………………………………………………………...25
Figure 5: Predictions of Latency to Post Given Social Rewards As A Function of Self-Report
Behavioral Automaticity Index (SRBAI) in Study 3…………………………………………..26
Figure 6: Predictions of Latency to Post After Platform Changes As A Function of Prior Post
Frequency in Study 3…………………………………………………………………………...27
Figure 7: Predictions of Latency to Post After Platform Changes As A Function of Self-Report
Behavioral Automaticity Index (SRBAI) in Study 3…………………………………………...28

vi
Abstract
Does regular posting on Facebook reflect a habit? Suggesting that it does, more frequent
posters reported posting more automatically. Further tests revealed that frequent posting
demonstrates hallmark patterns of habit: Study 1 showed insensitivity to reward in that frequent
and habitual users (but not infrequent/non-habitual ones) continued to post even when others
provided few social rewards (e.g., like emojis). Study 2 demonstrated habit dependence on
context cues in that frequent users were disrupted by changes in cues within the Facebook
platform that, in contrast, increased posts of occasional users. Finally, Study 3’s survey revealed
that frequent and habitual users were largely unaware of these effects. They predicted that they
would respond much like infrequent and non-habitual users by increasing their rate of posting
given rewarding feedback and design changes. In sum, habit provided new insights into online
behavior, and media use supported central tenets of classic habit theory.

Keywords: Social Media, Habit, Reward, Facebook, Motivation, Goals

vii
Preface: Statement of Relevance
We provide a novel demonstration that social media use reflects basic science principles
of habit formation and performance. In our first study, habitual Facebook posters’ posting
behavior was insensitive to the rewards of others’ reactions, whereas occasional posters were
reward-sensitive. Our second study used a 2007 Facebook platform design change to show that
strong social media habits depend on context cues and are disrupted when these change. A final
study demonstrated that habitual users have more difficulty than non-habitual ones in predicting
their responses to rewards and changes in cues. In summary, frequent Facebook users form
habits that guide their use of the site in ways that diverge from occasional users. These patterns
highlight the usefulness of habit science for understanding social media use and everyday social
interaction more broadly. In turn, social media provides novel behavioral indicators especially
suited to reveal the habit processes that guide human responding.

1
Facebook Habits: Rewards, Cues, and Automaticity
It might seem obvious that social media use can become a habit, but it is not clear that
posting on social media becomes habitual in the same way as do other responses. Habits are
context-response associations that develop in memory as people repeat rewarded responses in
stable contexts (Amodio, 2019; Wood & Runger, 2016). If social media use is habitual, then
rewards such as others’ likes, follows, and friends should be important primarily for building
habit associations in memory (de Wit, 2017). Once formed, media habits should be relatively
insensitive to outcomes. Habitual use would be triggered automatically by context cues, such as
phone alerts or locations, and would be disrupted when cues change (Verplanken & Orbell, in
press). In these ways, theorizing about habit identifies the conditions under which social media
use is likely to persist or be curtailed. In turn, social media use provides a novel setting in which
to test central tenets of classic habit theory.
The present research evaluated the habitual nature of posting on Facebook, the largest
and most widely-accessed social platform, with nearly 2.8 billion active users worldwide (We
Are Social, 2021). Our first study tested the role of rewards in creating social media habits.
Computational models have shown that the number of others’ likes and comments on Instagram
increases the frequency of subsequent posts (Lindström et al., 2021; see also Schnauber-
Stockman & Naab, 2019). A habit analysis makes the counterintuitive prediction that positive
reactions or comments on a post will be less important for habitual than non-habitual Facebook
posters (see Mazar & Wood, 2018). By examining these ideas with Facebook postings, we
provide a naturalistic, real-world alternative to laboratory tests of reinforcer devaluation, which
are designed to evaluate whether strong habits persist when rewards are no longer valued or
reward contingencies change. Although habits have not always proved insensitive to rewards in

2
such lab paradigms (de Wit et al., 2018), this insensitivity might be revealed in everyday
contexts that cue behaviors without the increased deliberation activated in lab experiments.
Our second study tested another basic habit feature, activation of performance by context
cues. Suggesting this dependence on cues, Vishwanath (2015) found that self-reported habitual
Facebook users responded automatically to phishing messages on Facebook, even when they
were highly motivated to protect online privacy. Dependency on cues should be especially
apparent when cues change within a platform site. Such effects were modeled in an experiment
that provided extensive initial practice with an online site, enabling participants to respond
quickly and accurately (Garaialde et al., 2020). However, after the design changed, responses
slowed and became less accurate, suggesting automatic responding was disrupted. Additionally
suggestive is an earlier finding that a Facebook platform change seemed to slow the posting rates
of users who posted frequently (Anderson & Wood, 2021). Platform changes thus can function
like habit discontinuities in daily life, in which changes in living contexts disrupt lifestyle habits
(Orbell & Verplanken, 2020). Our second study directly tests the disruptive effects of a posting-
relevant change in platform design.
Our final study tested a third aspect of habitual use: Facebook users’ understanding of the
reasons for their posting behavior. Habit memories are not accessible to conscious awareness in
the same way as are intentions and desires, and people responding out of habit preferentially
explain their behavior in agentic terms, as a result of their motives (Mazar & Wood, 2021). In so
doing, they embrace a broadly-held folk theory that most actions are intentional (Rosset, 2008).
In support, frequent social media users who gained a sense of belonging from a site did not
consider their behavior habitual, apparently attributing their use to feelings of belonging (Liu et

3
al., 2018). If habitual and occasional Facebook users rely on such folk theories, then they should
make similar predictions about their posting.
Study 1: Rewards Motivate Initial/Occasional Facebook Use
This study tested the relation between the latency between Facebook users’ posts over a
week-long period and rewards from others’ positive reactions. To assess habit strength, we
evaluated an antecedent of the development of habit, the frequency of past posting, and a
consequence, self-reported habit strength (Verplanken & Orbell, 2003). Our central prediction
was that users without habits would be motivated to post again by others’ “like” and “love”
reactions, whereas users with strong habits would be less motivated by such social rewards,
instead posting as cued by contexts. Thus, we predicted that analyses on latency to post again
would reveal an interaction between the number of positive reactions received and users’ habit
strength such that latency will decrease given positive reactions on the immediately prior post for
weak habit posters but not high.
Method
Design
For a week, we observed participants’ Facebook posts and recorded the timestamps on
each one in order to measure the exact latency between each post. In order to assess the extent to
which social rewards influence subsequent posts, coders in real time tracked others’ reactions
and comments as they were made. To our knowledge, this is one of the few live-monitoring
observational studies of social media use (see also Deters & Mehl, 2012; Lindstrom et al., 2019).
Participants

4
Our final sample of 122 participants was recruited through two sources: A snowball
sampling method identified 80 active Facebook posters from the psychology student pool at a
large Californian university, their family and friends, and posted advertisements placed around
campus; in addition, a paid online sample (Prolific) identified 42 twice-per week Facebook
posters. From these participants, we observed 1,907 Facebook posts. Despite a twice-per-week
requirement for participation, 57 additional participants were excluded because they did not post
on Facebook during the study period.
To determine the sample size sufficient to detect an estimated moderately sized effect
(Cohen's f
2
= .15), we used a power simulation tool, simr (Version 1.04; Green & MacLeod,
2016). A minimum sample size of 70 participants was estimated necessary to obtain .80 power.
A detailed description of these simulations and the R code used for the power analyses for all
studies can be found on OSF (https://osf.io/6ya5d/).
Procedure
During the initial session, participants provided background information and measures of
habit strength (see below) and added as a “friend” a blank Facebook profile (procedure adapted
from Deters & Mehl, 2012). Participants’ posts were observed for the following week via the
observer profile. Participant privacy was ensured by categorizing the observer profile’s “friends”
list as private.
To monitor comments and reactions to posts, two independent coders tallied Facebook
activity between 8AM and 11PM daily. All tallies were posted into shared, password-protected
spreadsheet documents (a “main” sheet and “backup,” enabling the collected data to be triple-
checked for accuracy by both on-duty research assistants as well as the lead author). Assistants
monitored accounts roughly every 15 minutes, while the lead author checked for consistency

5
between the main and backup sheets 4 times daily. During the final session, participants were
shown the list of comments on their own posts, and they coded these for valence (negative,
positive, neutral). Finally, participants were debriefed.
Measures
Self-Reported Habit Strength (SRHI). On a scale ranging from 1 (Agree) to 7
(Disagree), participants rated experienced automaticity of Facebook posting on the 12-item Self-
Report Habit Index (Verplanken & Orbell, 2003; alpha = .93). Items included, “when I post on
Facebook, I do so….”automatically,” “without thinking,” and “before I realize I am doing so.”
Past Posting Frequency. After opening their Facebook page and reviewing their own
posting history, participants reported their average number of weekly posts.
Reactions and Comments. Coders tallied the number of valenced emoji reactions to
each post as positive (like, love) or negative (anger, sad). Surprise (“wow”) and laugh (“haha”)
emojis were not coded, as their valence is unclear. To control for individual differences in
numbers of reactions, the counts of positive and negative reactions received on posts were
standardized and mean-centered using the overall means of these reactions per post in the
sample. The same results were obtained when our analyses were conducted on non-standardized
data. Posts on Facebook group pages and comments on others’ profiles were not included.
Positive comments and negative reactions and comments were also tallied. Perhaps due
to the low numbers of negatively-valenced emoji reactions (M = 0.15, SD = 0.83) and both
positive (M = 0.37, SD = 1.98) and negative comments (M = 0.05, SD = .39), preliminary
analyses (reported on OSF) revealed that these types of reactions had little effect, and we do not
discuss them further.

6
Between-Post Latency. Our primary dependent variable is the elapsed time in hours
between postings. Given that high-frequency posters had shorter latencies because they posted
more rapidly, the data were skewed by greater numbers of short-latency posts. Thus, analyses
were conducted on log transformed latencies.
Results
Correlations, means, and standard deviations are shown in Table 1. As predicted,
experienced automaticity as reflected in the self-report habit index (SRHI) was strongly
correlated with posting frequency, supporting that these are alternative measures of habit
strength. Between-post latency was negatively correlated with both posting frequency and self-
report habit (SRHI) measures, reflecting just that users who posted more often also, by
definition, posted more rapidly. Unexpectedly, greater latency was (weakly) associated with
more positive reactions. This association most likely reflects the structure of our sample, which
was limited to users who posted twice a week or more, and who may thus be insensitive to
reward.
Initial support for our hypothesis comes from the negative correlations between past post
frequency/SRHI and the number of positive reactions per post. Thus, users with stronger habits
received fewer positive reactions from others than users with weaker habits, suggesting that their
high use levels are not driven by numbers of rewards.
Table 1
Means, Standard Deviations, and Correlations from Study 1
M SD Latency Positive
reactions
SRHI

7
Latency (hours) 20.27 166.87
Positive
reactions
2.95 7.66 .13**
SRHI 3.17 1.39 -.15** -.19**
Self-reported
weekly posts
15.85 19.19 -.25** -.22** .59**

Note. Higher numbers reflect longer latencies between posts, more positive reactions
(emojis) to the prior post, stronger habits, and more self-reported weekly posts. All means and
standard deviations are non-standardized values, or raw counts. Correlations were computed
using the standardized values of each measure. Reaction counts are for the post immediately
prior to the focal post. For correlations, latency is the average amount of time elapsed between
posts in log hours. *p < .05 **p < .01

Analyses were conducted with multilevel (hierarchical) models in which the level 1
equation captures the association between standardized, mean-centered numbers of positive
reactions to the immediately prior post and log-normalized between-post latency (time until the
subsequent post). Level 2 models the interaction between habit strength (at the participant level)
and the lagged reactions (within-participant at the post level) predicting between-post latency.
(1) 𝐿𝑎𝑡𝑒𝑛𝑐𝑦ij = 𝛽0j + 𝛽1j𝑅𝑒𝑎𝑐𝑡𝑖𝑜𝑛𝑠 i-1j+ 𝜀 ji
(2𝑎) 𝛽 0j = 𝑦00 + 𝑦01𝑃𝑜𝑠𝑡𝐻𝑎𝑏𝑖𝑡 j×𝑅𝑒𝑎𝑐𝑡𝑖𝑜𝑛𝑠j + 𝑦02𝑃𝑜𝑠𝑡𝐻𝑎𝑏𝑖𝑡+ 𝑦03𝑅𝑒𝑎𝑐𝑡𝑖𝑜𝑛𝑠+ 𝑢0j
(2𝑏) 𝛽 1j = 𝑦10 + 𝑦11𝑃𝑜𝑠𝑡𝐻𝑎𝑏𝑖𝑡+ 𝑢1j
Habit Strength as Past Post Frequency
Estimation of the above model using past post frequency for habit strength revealed the
predicted interaction between number of positive reactions on the prior post and prior posting

8
frequency, 𝛽 = 0.38, 95% CI [0.15, 0.6], p = .001, df = 1840.73 (see Table 2 and Figure 1). To
graph the interaction, we plotted lines for the sample minimum of posting only two times per
week, and for daily posters, who posted on the site 7x per week. As predicted, among users who
posted infrequently, greater numbers of positive reactions to their immediately prior post reduced
the latency until the next post, 𝛽 = -0.32, 95% CI [-0.58, -0.06]. In contrast, users who posted
daily or more often did not increase their posting rates when they received greater numbers of
positively valenced reactions, 𝛽 = 0.05, 95% CI [-0.1, 0.2]. Thus, we see that the predicted
insensitivity to reward in the higher frequency group.
The overall model also revealed two main effects, for positive reactions and frequency,
reflecting that more positive reactions predicted slightly greater latency for all posters (see
explanation for correlations above), whereas greater self-reported weekly posts predicted slightly
lower latency (see Table 2 and Figure 1).
Table 2
Study 1 Multilevel Model Predicting Latency Between Posts From Past Posting Frequency
and Social Rewards
df 𝛽 p 95% CI
Intercept 47.0
1
1.43 .39 -0.46, 0.181
Positive reactions 1898
.38
0.43 .003 0.145, 0.7
Self-reported weekly posts 58.1
4
-0.96 < .001 -1.27, -0.65

9
Positive reactions x Self-
reported weekly posts
1840
.73
0.38 .001 0.15, 0.6

Note. Estimates are the non-standardized coefficients (𝛽) of the terms in the multilevel
model. The dependent variable is average between-post latency. Positive reactions, assessed for
each post, are the average number of positively-valenced emoji reactions on the immediately
prior post. Self-reported weekly posts is a participant-level variable. Degrees of freedom are
calculated using the Satterthwaite method.

Fig. 1

Plot of the Interaction in Study 1 Between Past Weekly Posts and Social Rewards Predicting
Latency Between Subsequent Posts

Note. Latency (log hours) between posts as a function of past frequency of weekly posts and
number of positive reactions to the immediately prior post. For the plot, low frequency users

10
were specified as posting 2x weekly, and high frequency ones as posting daily. Error bars
represent 95% CIs.

Habit Strength from Self-Reported Habit Index (SRHI)
The above model was also computed using the self-report habit index (SRHI) and again
revealed the predicted interaction between positively valenced reactions on the prior post and
SRHI, 𝛽 = 0.32 , 95% CI [0.18, 0.47], p < .001, df = 1727.08 (see Table 3). As can be seen in
Figure 2, greater numbers of positive reactions decreased latency to post among weaker habit
participants, 𝛽 = -0.26, 95% CI [-0.45, -0.05]; but had no effect for stronger habit ones, 𝛽 = 0.06,
95% CI [-0.09, 0.21]. Thus, we see the predicted effects for both weak and strongly habitual
posters. In this analysis, the number of positive reactions and the SRHI measure of habit strength
were also significant predictors of latency (see Table 3).
Table 3
Study 1 Multilevel Model Predicting Latency Between Posts From SRHI and Social Rewards
df 𝛽 p 95% CI
Intercept 67.03 0.31 .08 -0.38, 0.66
Positive
reactions
1824.
52
0.39 <
.001
0.18, 0.60
Self report
habit index (SRHI)
88.79 -0.42 .00
8
-0.74, -0.11
Positive
reactions x SRHI
1727.
08
0.32 <
.001
0.18, 0.47

11
Note. Estimates are non-standardized coefficients (𝛽) in a multilevel model predicting
between-post latency from number of positive reactions (emojis) on the immediately prior post,
self-reported habit strength (SRHI, participant-level), and the interaction between SRHI and
positive reactions. Degrees of freedom are calculated using the Satterthwaite method.

Fig. 2

Plot of the Interaction in Study 1 Between Self-Report Habit Index (SRHI) and Social
Rewards Predicting Latency Between Subsequent Posts

Note. Latency (log hours) between posts as a function of habit strength (SRHI) and number
of positive reactions to the immediately prior post. Weak habits line corresponds to standardized
SRHI scores for users posting 2x per week. Strong habits line corresponds to standardized SRHI
scores for users posting 7x per week. Error bars represent 95% CIs.

Single and Multiple Lag Results

12
Because each new post latency is driven by reactions that occurred before that post, our
reaction counts could be calculated by tallying the number of reactions on the prior 1, 2, and 5
post(s). We used each of these delays to calculate 3 different models. The single-post lag model
(reflecting the quantity of reactions received on the immediately prior post) should provide the
strongest evidence of reward effects, and rewards might dissipate at longer delays. However, if
our results reflect alternative, perhaps more enduring influences on online behavior, then we
would expect the effects with single lag models to maintain at longer lags.
As reported in the OSF supplement (https://osf.io/6ya5d/), a 2-post lag model using the
same predictors showed similar effects to the ones we report in the text. However, the 5-post lag
model did not yield the predicted interaction effect, suggesting that the 1- and 2- post lag models
reflect the causal effects of rewards on immediately subsequent posts that decay over time.
Discussion
The findings of Study 1 confirmed our prediction that people who posted infrequently
and were weakly habitual Facebook users, positive reactions from others on an immediately
prior post decreased latency until the next post. In contrast, high-frequency and strongly habitual
posters displayed the predicted reward insensitivity.
In demonstrating that strong habit performance is relatively insensitive to rewards, Study
1 supports a central tenet of habit theory. Note that insensitivity has not always been obtained in
laboratory research using the reinforcer devaluation paradigm adapted from animal learning (de
Wit et al., 2018). In lab contexts, people might readily deploy attentional and goal systems to
override unwanted habits. Perhaps they are less able to do so in everyday settings characterized
by multiple demands on processing capacity. Another explanation is that habit insensitivity

13
reflects the quantity or intensity of rewards, and does not occur with the undesirable outcomes in
reinforcer devaluation studies.
Study 2: Cues Disrupt Habitual Users of Facebook
The second study assessed another canonical indicator of habits: disrupted responding
when context cues change. Specifically, we evaluated a platform change designed to enhance
Facebook user experience. However, to the extent that Facebook use is habitual, changes in
platform cues could have the unintended consequence of disrupting the automaticity of habitual
posters. Specifically, we predicted an interaction between timing of the platform change and
prior posting frequency such that more frequent posters would increase in latency, whereas
occasional posters might decrease latency, at least to the extent that the design change had its
intended effect. Our analysis builds on an earlier demonstration with a different platform change
that disrupted posting habits (Anderson & Wood, 2021). However, this earlier test failed to rule
out artifactual influences such as regression to the mean.
Method
Design
We evaluated a previously untested Facebook platform change on wall postings using an
existing data set (Vishwanath et al., 2009). Specifically, we studied how wall posting rates were
influenced the month before versus the month after a platform change on April 20, 2007. This
change was designed to encourage use by presenting a greater amount of friends’ content at the
top of a user’s newsfeed along with the addition of the phrase, “USERNAME is..,” next to the
status update bar. Given the historical nature of this dataset, users were posting directly within
the Facebook platform rather than sharing posts from outside the platform. From this dataset, we
included N = 5,319 Facebook users, who had posted at least once during a 6-month period prior

14
to the change (September 20, 2006 to March 20, 2007). These users made 41,199 wall posts
between March 20, 2007 and May 20, 2007.
Measures
Latency of Facebook Wall Posts. Posting rate was measured as the log latency of time
between posts during the month before the platform change (March 20, 2007 to April 20, 2007)
as well as latency during the month after the change (April 20, 2007 to May 20, 2007). This
measure was transformed to a natural log scale to mitigate the strong skew in observed posting
behavior, as more rapid posts are by definition also more frequent posts. This variable is
structured identically to the dependent variable in Study 1.
Prior Posting Frequency. We estimated habit strength from prior posting frequency,
using the posting behavior of each user (quantity of posts) during the 6 months prior to the start
of assessment of the impact of the change (i.e., September 20, 2006 to March 20, 2007). In so
doing, we build on Study 1’s finding that prior posting frequency was substantially correlated
with perceived habit strength (SRHI).
Self-Post vs. Other-Post. In an exploratory analysis, we examined whether the platform
change had differential effects on posts to a users’ own walls as opposed to others’ walls. Given
that the change made the content of Facebook friends more visible, it should especially disrupt
posting on others’ walls. Consistent with this possibility, the predicted interaction held in the
analysis on wall posts to others but not the analysis on posts to one’s own wall. However, this
failure to find an effect with self posts could also be due to low power, given the small sample
size of self-posts (N = 294). Thus, we do not present these results in further detail here. These
results can be found in the OSF supplement (https://osf.io/6ya5d/).

15
Results
As can be seen in Table 4, the prior frequency of posting was negatively correlated with
the latency between posts, indicating simply that users who posted more often in the six months
prior to our assessment of platform change also posted more often during the two month-long
assessment period.
Table 4
Study 2 Means, Standard Deviations, and Correlation
M SD Frequency of posts in
the prior 6 months
Frequency of posts in
prior 6 months
31.28 35.53
Latency between posts
(seconds) during month
before and after platform
change
77976.
34
1792592.
02
-.25**

Note. Higher numbers reflect more posts and longer time elapsed between posts. The
correlation was computed with the ln(seconds) to reduce skew.

Posting Latency Before vs. After Change in Platform Cues
A multilevel regression model predicted the log latency between users’ posts from an
interaction between their frequency of posting in the prior six months (habit strength) and a
variable indicating whether a given post occurred in the month immediately before vs. after the
platform change. The expected interaction between habit strength and the change in platform
design was significant (see Table 5), 𝛽 = 0.003, df = 40,230, p < .001. To illustrate the

16
interaction, we performed a spotlight analysis on at-least-once-per-week posters (i.e., 26 posts
over the 6 study months) and at-least-daily posters (i.e., 180 posts over 6 months)--the same
criteria used in Study 1 to define strong and weak habits. Consistent with our predictions, the
latency decreased for posters with weak habits was, 𝛽 = -0.10, CI [-0.15, -0.05], p < .001,
whereas it increased for those with strong habits, 𝛽 = 0.36, CI [0.58, 0.14], p < .001. These
associations are displayed in Figure 3, below.

Table 5
Results of Multilevel Regression Model Predicting Latency Between Facebook Posts As A
Function of A Facebook Platform Change and Past Posting Frequency: Study 2
df 𝛽 p 95% CI
Intercept 8,95
9
12.
91
<
.001
12.84,
12.98
Frequency of posts in the prior 6
months (habit strength)
3,70
5
-
0.028
<
.001
-0.03 , -
0.025
Timing of change (0 = before, 1=
after)
41,0
40
-
0.181
<
.001
-0.25, -0.11
Frequency of prior posts x Timing
of change
40,2
30
0.0
03
<
.001
0.001,
0.004

17
Note. Estimates are the non-standardized coefficients (𝛽) of the terms in the multilevel model
predicting latency between posts (log seconds). Number of past posts is a participant-level
variable, and timing of the platform change is a group-level variable (value of 0 before April 20.
2007 at 12am, value of 1 after this date). Degrees of freedom were calculated using the
Satterthwaite method.

Two additional effects also emerged from the analysis (see Table 5). First, latency was
negatively associated with habit strength, reflecting the expected shorter latency between posts
among more frequent users. Second, latency was negatively associated with the change itself,
suggesting that overall, the change reduced users’ latency (i.e., increased their posting rates on
aggregate) as Facebook had originally intended.
Fig. 3

Plot of Latency Between Facebook Posts As A Function of the Timing of A Platform Change
and Past Posting Frequency; Study 2

18

Note. The criterion is posting latency (log seconds), with higher numbers reflecting greater
latency between users’ Facebook posts. Predictors are frequency of past posts, with strong habits
defined as 180 posts during the 6 months prior to the assessment, and weak habits as 26 posts.
Latency between posts is represented for a month prior to a Facebook platform design change
and the month after. Error bars represent 95% CIs.

Testing Artifacts: Comparing Posting Latency Before vs. After Randomly Chosen Dates
To test whether our findings are due to artifacts such as regression to the mean, we
selected an arbitrary date of one month after our assessment, May 20, 2007, and evaluated
posting latency before and after that date. Suggesting that the above effect is not artifactual, no
significant interaction emerged between users’ prior posting frequency and the month before vs

19
after this date, 𝛽 = 0.001, df = 41,530, p = .16. When we extended this analysis to include the day
after, May 21, 2007, and the day before, May 19, 2007, again no significant interaction emerged,
𝛽 = 0.001, df = 41,500, p = .12, and 𝛽 = 0.001, df = 41,550, p = .14, respectively. The full model
table and plot are available in the OSF supplement (https://osf.io/6ya5d/).
Discussion
In summary, our test of platform changes revealed the vulnerability of strongly habitual
users to changes in platform design. Although this change was broadly effective in boosting
consumers’ overall wall posting, this overall success hides an important story illuminated by a
habit-based analysis: Cue disruptions impede automated responses and orient people to respond
more thoughtfully. If people rely on intentions to act, then they will deliberately decide to
continue some behaviors and discontinue others (Wood et al., 2005).
Importantly, when we tested change in posting latencies on three randomly-selected
dates, we found that users did not change their posting behavior given no platform change. Thus,
artifacts such as regression to the mean do not account for platform effects. Also noteworthy is
that habitual posters slowed their posting rates after the platform change, despite that it speeded
up posting across all Facebook users. Nonetheless, these multiple influences on posting rates
might underlie the somewhat smaller changes in posting latency in Study 2 than observed in
Study 1. The smaller effect in this second study could also be due to our design, which tracked
the influence of platform effects across a full month, and thus may have allowed platform design
effects to dissipate especially for frequent users.
Our current findings are an important demonstration of habit discontinuity because past
investigations have not targeted changes in the specific cues responsible for habit disruption and
instead have relied on broader shifts in living contexts such as moving house or starting a new

20
job (Verplanken & Orbell, 2020). By identifying changes in specific platform cues that promote
or impede posting, we provide especially strong support for discontinuity effects.
Study 3: Predictions about Posting Given Rewards and Platform Changes
The final study assessed a third indicator of habitual use: understanding of what
influences behavior. People who post frequently have many opportunities to observe their
behavior and as a result might accurately understand the determinants of their posts. However, to
the extent that frequent posting becomes habitual, frequent posters may be unaware of the cues
that trigger this behavior (Mazar & Wood, 2021; Wood et al., 2017). Without an attitude or
belief in mind to account for their behavior, habitual users might instead rely on lay theories
about why people use Facebook. Given that such lay accounts give preference to agentic,
intentional explanations (Rosset, 2008), habitual Facebook users might make similar predictions
as occasional ones. These could draw on the reasons to initially use Facebook, such as desires to
connect with others and garner social rewards.
If all users rely on lay theories of intentionality, then they are uniformly likely to predict
that they will be influenced to post more by others’ positive reactions as well as by platform
design changes intended to increase posting. In comparison with the behavioral results from our
first two studies, then, strongly habitual posters are likely to be less accurate in their predictions
than more occasional posters.
Method
Participants were 243 Facebook users recruited via Amazon’s MTurk CloudResearch
panels who were U.S.-based English speakers with a Facebook account. Our a-priori power
analysis was performed in G*Power, finding that for a moderate effect size, Cohen's f
2
= .15, a
minimum of 55 participants was needed to achieve power greater than 80%. An additional 6

21
users were removed from the data for missing crucial data points or unfinished surveys, and 2
other users were eliminated for extreme outlier values. The study method and plan were pre-
registered (https://osf.io/dfcy6/).
Measures
Past Frequency of Posting on Facebook. As in both prior studies, the past frequency of
weekly posting was one indicator of habit strength.
Self-Report Behavioral Automaticity Index (SRBAI; Gardner et al., 2012).
Participants indicated habit strength using this subscale of the SRHI (Verplanken & Orbell,
2003), which is composed of four items: “Posting on Facebook is something...I do
automatically”, “...I do without having to consciously remember”, “...I do without thinking”, and
...I start doing before I realize I’m doing it (1= Strongly disagree; 7 = Strongly agree). This
scale was used instead of the full SRHI (as in Study 1) in order to reduce participant load.
Predicted Response to Social Rewards. Participants indicated whether receiving
positive or negative reactions from others would affect their time to post again: “Imagine that
you get positive [negative] reactions or comments from others on a Facebook post. How would
this impact your next posts?” (1 = I wouldn’t post again for a long time; 4 = it wouldn’t change
how I posted; 7 = I would post again almost immediately). As in Study 1, we focus only on
positive reactions. A correlation table including negative reactions can be found in the OSF
Supplement (https://osf.io/dfcy6).
Predicted Response to Change in Platform Cues. Participants indicated whether
changes in platform cues (as in Study 2), would affect their time to post again: “If Facebook put
your friends’ posts closer to the top of your newsfeed and added a prompt to post what you’re

22
doing right now, how would this impact your next posts?” (1 = I wouldn’t post again for a long
time; 4 = it wouldn’t change how I posted; 7 = I would post again almost immediately).
Results
As can be seen in Table 6, predictions about rewards and cue changes were associated,
indicating that people who believed their posting rate would increase given rewards also believed
it would increase given the platform changes. Additionally, perceived automaticity of posting
(SRBAI) and past posting frequency were positively correlated, as in Study 1, suggesting they
are both measures of habit strength.
In initial support of our expectation that strongly habitual users would rely on the same
lay theories to predict their reactions as less habitual users, past posting frequency was not
correlated with users’ predicted responses to rewards. Furthermore, when habit strength was
correlated with predictions about rewards and platform changes, more habitual users predicted
that they would be more strongly influenced by rewards and platform changes.
Table 6
Study 3 Means, Standard Deviations, and Correlations
M SD Predicted
response to
rewards
Predicted
response to
cues
Past posting
frequency
Predicte
d response
to rewards
4.
89
1.1
3

23
Predicte
d response
to cues
4.
08
0.6
2
.30**
Past
posting
frequency
14
.41
9.8
2
.11 .13*
SRBAI 2.
66
1.5
5
.27** .19** .38**

Note. Predicted response to rewards and cues are both on 1-7 scales with higher scores
reflecting less time (latency) between posts (i.e., more rapid posting). Past posting frequency is a
raw count of users’ average weekly posts. Self-report behavioral automaticity index (SRBAI) is a
4-item, 5-point scale, with higher numbers reflecting greater automaticity (habit strength). Full
table including all measures is available in the OSF supplement (https://osf.io/6ya5d/). * p < .05,
** p < .01

Habit Strength and Anticipated Latency to Post
Response to social rewards. A regression model revealed that, as expected, prior posting
frequency was not a significant predictor of participants’ anticipated responses to social rewards,
𝛽 = 0.01, t(241) = 1.69, p = .09. This analysis was replicated using self-reported automaticity
(SRBAI) as the measure of habit strength and revealed that participants with stronger habits
predicted that they would post faster if they got more social rewards, 𝛽 = 0.19, t(241) = 4.44, p <
.001.
Response to platform change. A regression model revealed that prior posting frequency
influenced anticipated responses to platform changes, 𝛽 = 0.008, t(241) = 2.07, p = .04, such that
participants who had posted more frequently in the past anticipated that they would post again
faster given the design change. In a separate regression, participants with stronger habits

24
(SRBAI) also predicted that they would post faster to design changes, 𝛽 = 0.08, t(241) = 3.08, p
= .002.
Accuracy of Predictions
We compared the predictions given by participants in Study 3 with the actual responses
given by comparable participants in Studies 1 and 2. Specifically, we computed t-tests to
examine whether posters’ predictions about their likely responses to rewards and platform
changes were significantly different from the scale midpoint of 4, “I would not change my
posting behavior.”
Social rewards. Habitual posters would accurately mirror Study 1 findings if their
predictions were close to the scale midpoint of no change, whereas nonhabitual posters would be
accurate if their predictions were a decrease in latency between posts. However, when analyses
were conducted with past post frequency as the indicator of habit strength, the mean for the
highly frequent posters was significantly greater than 4, M = 5.00, 95% CI = [4.79, 5.21], t(111)
= 9.23, p < .001, indicating that they (incorrectly) predicted they would post faster after getting
more positive reactions from others (see Figure 4). The low frequency group’s prediction that
they would post faster given more positive reactions also differed significantly from the scale
midpoint, but in this case accurately reflected Study 1 results, M = 4.80, 95% CI = [4.61,4.99],
t(130) = 8.23, p < .001.
Fig. 4
Predictions of Latency to Post Given Social Rewards As A Function of Prior Post Frequency
in Study 3

25

Note. Participants’ predictions about their speed of posting after social rewards, reported on a
7-point scale on which higher numbers represent faster posting. Error bars indicate the 95%
confidence interval (CI) for each group mean. Weak/strong habit groupings are based on a mean
split of standardized prior weekly posts (habit strength).

We repeated these analyses using habit strength as reflected in the self-report behavioral
automaticity index (SRBAI, see Figure 5). Again, the mean for the strongly habitual posters was
greater than the scale midpoint, M = 5.21, 95% CI = [4.99, 5.44], t(102) = 10.62, p < .001,
indicating that they (incorrectly) believed they would post faster given positive reactions from
others. The weak habits group was also higher than the scale midpoint, M = 4.66, 95% CI =
[4.48, 4.83]t, t(139) = 7.45, p < .001, which actually corresponded with behavior in Study 2.
Fig. 5

26
Predictions of Latency to Post Given Social Rewards As A Function of Self-Report
Behavioral Automaticity Index (SRBAI) in Study 3

Note. Participants’ predictions about their speed of posting after social rewards, reported on a
7-point scale on which higher numbers represent faster posting. Error bars indicate the 95%
confidence interval (CI) for each group mean. Weak/strong habit groupings are based on a mean
split of standardized SRBAI (habit strength) measures.

Change in platform design. Habitual posters would accurately mirror Study 2 if their
predictions were below the scale midpoint of no change, indicating that they would increase
latency between posts, whereas nonhabitual posters would be accurate if their predictions were
above the scale midpoint, indicating they would decrease latency between posts. However,
analyses using mean frequency of past posts as an indicator of habit strength revealed that more
frequent posters predicted (inaccurately) a higher score than the scale midpoint, and reported that
they would post more quickly, M = 4.15, 95% CI = [4.03, 4.27], t(111) = 2.48, p = .01 (see

27
Figure 6). In contrast, low frequency posters’ predictions did not differ from the midpoint,
indicating (again inaccurately) that the change would have no effect, M = 4.02, 95% CI = [3.92,
4.13], t(130) = 0.43, p = .66.
Fig. 6
Predictions of Latency to Post After Platform Changes As A Function of Prior Post
Frequency in Study 3

Note. Participants’ predictions about their posting behavior after platform changes, reported
on a 7-point scale on which higher numbers represent faster posting. Error bars indicate the 95%
confidence interval (CI) for each group mean. Weak/strong habit groupings are based on a mean
split of standardized prior weekly posts (habit strength) measures.

We replicated these analyses using the self-report automaticity measure (SRBAI) as an
indicator of habit strength. Again, strongly habitual posters (incorrectly) anticipated that they

28
would post again more quickly after the platform change, differing significantly from the scale
midpoint, M = 4.20, 95% CI = [4.07, 4.34], t(102) = 3.06, p = .002 (see Figure 7). In contrast,
weak habit participants predicted (also incorrectly) that they would not be influenced by the
design change, and their ratings did not differ from the scale midpoint, M = 3.99, 95% CI =
[3.90, 4.09], t(139) = -0.15, p = .88.
Fig. 7
Predictions of Latency to Post After Platform Changes As A Function of Self-Report
Behavioral Automaticity Index (SRBAI) in Study 3

Note. Participants’ predictions about their speed of posting after a platform change, reported
on a 7-point scale on which higher numbers represent faster posting. Error bars indicate the 95%
confidence interval (CI) for each group mean. Weak/strong habit groupings are based on a mean
split of standardized SRBAI (habit strength) measures.

29

Discussion
This study assessed Facebook users' predictions about how they would act in the settings
captured in our first two studies of (a) others responding positively to posts and (b) a platform
change designed to promote posting. Overall, the findings reveal a striking level of inaccuracy.
All participants expected that they would post faster again when they got more likes and loves on
a prior post. Compared with the actual findings from Study 1, this prediction was accurate for the
weakly habitual users but not the strongly habitual ones. Given a platform change, strongly
habitual posters again predicted they would post more quickly, whereas weakly habitual ones
anticipated no change in latency to post. Both of these predictions are inaccurate given Study 2
findings. It is striking that these patterns were highly consistent across our two very different
measures of habit strength.
The lack of correspondence between the actual posting patterns (Studies 1 and 2) and
participants’ predictions in Study 3 is consistent with participants relying on implicit theories
about what drives their behavior and not reporting their personal observations. Especially for
social rewards, all Facebook users appear to be relying on a common lay theory to anticipate
their behavior. It seems that habitual posters are largely unaware of the habit mechanisms that
drive their behavior. Instead, they seem to assume that their behavior is a more intentional
response to others’ rewards and to the purpose of the platform design change, which was to
increase posting.
General Discussion
Across three studies of Facebook postings, we provide support for existing habit theory
in a novel domain and demonstrate that frequently repeated social media posting is a habitual,

30
automatic behavior. Study 1 showed that high-frequency and habitual Facebook posting is
insensitive to numbers of social rewards in the form of others’ positive reactions. Study 2
demonstrated that high-frequency Facebook posting depends on context cues in the platform
design, in that posting was impeded among habitual users when design cues changed. Study 3
compared users’ predictions about posting with these actual effects in Studies 1 and 2. In contrast
with the actual posting rates, both habitual and non-habitual posters seemed to rely on intentional
theories about their behavior that, especially in the case of habitual posters, led to inaccurate
predictions about their responses. Habitual users anticipated acting much like non-habitual ones,
being driven to post again quickly when others gave positive rewards and when the platform
design changed to facilitate posting. Taken together, these three studies identify novel patterns in
Facebook posting that are best explained through the lens of modern theories of habit.

31
References
Anderson, I. A., & Wood, W. (2020). Habits and the electronic herd: The psychology behind
social media’s successes and failures. Consumer Psychology Review.
Amodio, D. M. (2019). Social cognition 2.0: An interactive memory systems account. Trends in
Cognitive Sciences, 23(1), 21-33.
Bar-Anan, Y., Wilson, T. D., & Hassin, R. R. (2010). Inaccurate self-knowledge formation as a
result of automatic behavior. Journal of Experimental Social Psychology, 46(6), 884-894.
Deters, F. große, & Mehl, M. R. (2013). Does posting Facebook status updates increase or
decrease loneliness? An online social networking experiment. Social Psychological and
Personality Science, 4(5), 579–586. https://doi.org/10.1177/1948550612469233
de Wit, S. (2017). Control of behavior by competing learning systems. In T. Eger (Ed.), The
Wiley handbook of cognitive control (pp. 190-206). Wiley & Sons.
de Wit, S., Kindt, M., Knot, S. L., Verhoeven, A. A., Robbins, T. W., Gasull-Camos, J., ... &
Gillan, C. M. (2018). Shifting the balance between goals and habits: Five failures in
experimental habit induction. Journal of Experimental Psychology: General, 147(7),
1043-1065. http://dx.doi.org/10.1037/xge0000402
Duckworth, A. L., Gendler, T. S., & Gross, J. J. (2016). Situational strategies for self-control.
Perspectives on Psychological Science, 11(1), 35-55.
Garaialde, D., Bowers, C. P., Pinder, C., Shah, P., Parashar, S., Clark, L., & Cowan, B. R.
(2020). Quantifying the impact of making and breaking interface habits. International
Journal of Human-Computer Studies, 142, 102461.
Gardner, B., Abraham, C., Lally, P. et al. Towards parsimony in habit measurement: Testing the
convergent and predictive validity of an automaticity subscale of the Self-Report Habit
Index. International Journal of Behavioral Nutrition & Physical Activity, 9, 102 (2012).
https://doi.org/10.1186/1479-5868-9-102
Green, P., & MacLeod, C. J. (2016). SIMR: an R package for power analysis of generalized
linear mixed models by simulation. Methods in Ecology and Evolution, 7(4), 493-498.
Hardwick, R. M., Forrence, A. D., Krakauer, J. W., & Haith, A. M. (2019). Time-dependent
competition between goal-directed and habitual response preparation. Nature Human
Behaviour, 3(12), 1252-1262.
Itzchakov, G., Uziel, L., & Wood, W. (2018). When attitudes and habits don’t correspond: Self-
control depletion increases persuasion but not behavior. Journal of Experimental Social
Psychology, 75, 1-10.

32
Lin, P. Y., Wood, W., & Monterosso, J. (2016). Healthy eating habits protect against
temptations. Appetite, 103, 432-440.
Lindström, B., Bellander, M., Schultner, D. T., Chang, A., Tobler, P. N., & Amodio, D. M.
(2021). A computational reward learning account of social media engagement. Nature
Communications, 12(1), 1-10.
Liu, Q., Shao, Z., & Fan, W. (2018). The impact of users’ sense of belonging on social media
habit formation: Empirical evidence from social networking and microblogging websites
in China. International Journal of Information Management, 43, 209-223.
Mazar A., Wood W. (2018) Defining habit in psychology. In B. Verplanken (Ed.), The
psychology of habit (pp. ). Springer, Cham. https://doi.org/10.1007/978-3-319-97529-0_2
Neal, D. T., Wood, W., Labrecque, J. S., & Lally, P. (2013). How do habits guide behavior?
Perceived and actual triggers of habits in daily life. Journal of Experimental Social
Psychology, 48(2), 492–498. https://doi.org/10.1016/j.jesp.2011.10.011
Orbell, S., & Verplanken, B. (2010). The automatic component of habit in health behavior: Habit
as cue-contingent automaticity. Health Psychology, 29(4), 374.
Orbell, S., & Verplanken, B. (2020). Changing behavior using habit theory. In Hagger M.,
Cameron L., Hamilton K., Hankonen N. & Lintunen T. (Eds.), The Handbook of
behavior change (pp. 178–192). Cambridge University Press. DOI:
10.1017/9781108677318.013
Rosset, E. (2008). It’s no accident: Our bias for intentional explanations. Cognition, 108(3), 771–
780. https://doi.org/10.1016/j.cognition.2008.07.001
Verplanken, B., & Orbell, S. (2003). Reflections on past behavior: A self‐report index of habit
strength 1. Journal of Applied Social Psychology, 33(6), 1313-1330.
Verplanken, B., & Orbell, S. (in press). Attitudes, Habits and Behavior Change. Annual Review
of Psychology. https://doi.org/10.1097/00005053-196001000-00018
Vishwanath, A. (2015). Habitual Facebook use and its impact on getting deceived on social
media. Journal of Computer-Mediated Communication, 20(1), 83-98.
Vishwanath, B., Mislove, A., Cha, M., & Gummadi, K. P. (2009). On the evolution of user
interaction in Facebook. In Proceedings of the 2nd ACM workshop on online social
networks, USA, pp. 37–42. https://doi. org/10.1145/1592665.1592675
We Are Social, & Hootsuite, & DataReportal. (January 27, 2021). Most popular social networks
worldwide as of January 2021, ranked by number of active users (in millions) [Graph]. In
Statista. Retrieved March 12, 2021, from https://www-statista-
com.libproxy1.usc.edu/statistics/272014/global-social-networks-ranked-by-number-of-
users/

33
Wood, W. (2017). Habit in personality and social psychology. Personality and Social
Psychology Review, 21(4), 389–403. https://doi.org/10.1177/1088868317720362
Wood, W., Mazar, A., & Neal, D. (2021). Habits and Goals in Human Behavior: Separate but
Interacting Systems.
Wood, W., & Rünger, D. (2016). Psychology of habit. Annual Review of Psychology, 67.
Wood, W., Tam, L., & Witt, M. G. (2005). Changing circumstances,disrupting habits. Journal of
personality and social psychology, 88(6), 918.

Asset Metadata

Creator Anderson, Ian A. (author)

Core Title Facebook habits: rewards, cues, and automaticity

Contributor Electronically uploaded by the author (provenance)

School College of Letters, Arts and Sciences

Degree Master of Arts

Degree Program Psychology

Degree Conferral Date 2021-08

Publication Date 07/29/2021

Defense Date 07/27/2021

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag cues,facebook,goals,Habits,Learning,media,OAI-PMH Harvest,reward learning,rewards,social media,Technology

Format application/pdf (imt)

Language English

Advisor Wood, Wendy (committee chair), Read, Stephen (committee member), Schwarz, Norbert (committee member)

Creator Email ianaxelanderson@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC15669264

Unique identifier UC15669264

Legacy Identifier etd-AndersonIa-9937

Document Type Thesis

Format application/pdf (imt)

Rights Anderson, Ian A.

Type texts

Source University of Southern California (contributing entity), 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 uscdl@usc.edu

Abstract (if available)

Abstract Does regular posting on Facebook reflect a habit? Suggesting that it does, more frequent posters reported posting more automatically. Further tests revealed that frequent posting demonstrates hallmark patterns of habit: Study 1 showed insensitivity to reward in that frequent and habitual users (but not infrequent/non-habitual ones) continued to post even when others provided few social rewards (e.g., like emojis). Study 2 demonstrated habit dependence on context cues in that frequent users were disrupted by changes in cues within the Facebook platform that, in contrast, increased posts of occasional users. Finally, Study 3’s survey revealed that frequent and habitual users were largely unaware of these effects. They predicted that they would respond much like infrequent and non habitual users by increasing their rate of posting given rewarding feedback and design changes. In sum, habit provided new insights into online behavior, and media use supported central tenets of classic habit theory.