University of Southern California Dissertations and Theses

Responding to harm

(USC Thesis Other)

Responding to harm

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Responding to Harm
b y
Da vid J. Clark
a Dissertation Presented to the
F A CUL T OF THE USC GR ADUA TE SCHOOL
UNIVERSITY OF S OUTHERN CALIFORNIA
in Partial Fulfillment of the
Requirements for the Degree
DOCT OR OF PHILOSOPHY
(PHILOSOPHY)
August 2023
Cop yright 2023 Da vid J. Clark
Ackno wle dgments
There are so man y pe op le to thank. I’m sure I’v e f aile d to ackno wle dge man y of them.
I suppose I should start at the beginning, when I first enc ountere d p hilosop h y as a c adet at
the U.S. Military Ac adem y at W est P oint. I ha v e Fr. Timoth y V alentine, m y first p hilosop h y
instructor, to thank f or m y introduction to the subje ct. He had me imme diately hoo ke d, and
I ha v en’t be en ab le to wrestle fre e of p hilosop h y sinc e (despite m y best eff orts).
I c omp lete d m y undergraduate e duc ation at Bio la Univ ersity , where I had the p leasure
of learning p hilosop h y under the wings of (to m y mind) some of the v ery best undergrad
teachers and mentors: Da vid Cioc chi, T om Crisp , Chris Franklin, Kristen Irwin, and Gregg
T enElshof. E v ery one of them supporte d me in more w a ys than I c ould hope to tell, but I am
espe cially grateful to T om Crisp , who has remaine d a v ery dear friend and brother, and who
I will miss terrib ly when it c omes time to lea v e Calif ornia.
The ne xt chapter of m y lif e too k p lac e outside of ac ademia. M ari and I w elc ome d our first
child (M ap le), and w e too k up one of the only prof essions w e c ould think of that has w orse
“return on in v estment” than ac ademic p hilosop h y: f arming. W e lease d and manage d a f arm
ii
in c entral V ermont, which left little time f or much else. But when I f ound time to get a w a y
from the f arm, I alw a ys f ound a w elc oming p hilosop hic al interlocutor in T erenc e Cune o ,
who kept the fire lit. His f amily w as a great support to ours, and T erenc e w as a sourc e of
much enc ouragement as a serious p h ysic al injury too k me a w a y from f arming a nd ev entually
b ack into ac ademia. (I also ne e d to thank Andrew Bailey and P eter V an Elswyk f or their
support throughout this transition. I w ould ha v e be en lost without their dire ction, and m y
writing samp le w ould ha v e be en ev en lousier without their detaile d c omments.)
Ne xt stop: Northern Illinois Univ ersity . I don’t miss y ou, DeKalb , I’m sorry . But I do
miss the w onderful Philosop h y dep artment at NIU, which w as full of rock star mentors and
teachers. I miss talking logic and b aseb all with Da vid Buller, trading v egetab les f or Alicia
Finch’s metap h ysic al e xpertise, grabbing cheap be ers and cheaper pizzas with Carl Gillett,
and haunting Stev e Daskal ’s offic e hours. But abo v e all I miss Craig and Laura W armke, who
opene d their home to M ari, M ap le, and I, and whose c ontinue d friendship I am thankful f or.
Outside of the f aculty at NIU, I am espe cially grateful f or the friendship of Elyse Oakley ,
Ali Aenehzodae e, and Tim Buchanan. I’v e had the great f ortune of sta ying close with Tim
as he also lande d in Los Angeles (though, regrettab ly , at UCLA), but the great misf ortune of
being sep arate d from Elyse and Ali. This is prob ab ly good f or m y quantity of sle ep , but it’s
v ery b ad f or m y quantity of fun.
W e made the mo v e to USC in 2017 (ha ving in the meantime adde d y et another little
iii
one to the clan, Nehemiah). I lande d in the best c o hort I c ould ha v e aske d f or with Jaime
Castillo-Gambo a, No ah Gordon, Jasmine Gunkel, M ahmoud Jallo h, M adiha Hamdi, An-
thon y Nguy en, and Andrew Stew art. They ha v e each be en w onderful c o lleagues. I am es-
pe cially thankful to M adiha and Andrew , f or their support and friendship . Outside of m y
c o hort I ne e d to shout out Irene Bosc o , Rebe c c a Carlson, Jen F oster, Zach Goodsell, Brian
Haas, and Frank Hong — who ha v e on man y oc c asions b lesse d me with their support, c are,
and brain po w er. I am espe cially thankful to Laura Gurskey , who has be en a better friend
than I deserv e, and who has surviv e d man y rounds of the Herculean task of w atching m y
brood (which has numbere d thre e sinc e little W endell ’s arriv al in 2019).
The f aculty in the USC Philosop h y Dep artment has be en brilliant and supportiv e at ev ery
turn. Among those whose brillianc e and support ha v e most imp acte d me are Jo hn Ha wthorne,
Stev e Finla y , Jeff Russell, and Sc ott So ames. I’m thankful to Jo hn f or making p hilosop h y so
damn fun. I’m thankful to Jeff f or man y c omments on m y written w ork, man y rich discus-
sions, and f or his v aluab le and detaile d guidanc e in job market prep aration. I’m thankful to
Stev e f or his early enc ouragement, f or pro viding more f e e db ack on m y w ork than an y other
person outside m y c ommitte e, and f or giving me the opportunity to c o llaborate with him
as a research assistant. I’m thankful to Sc ott f or putting up with me twic e as his teaching
assistant, f or his generous mentorship and support while I w orke d as his research assistant,
and f or our man y fun lunch discussions. Others to whom I o w e much include Shiev a Klein-
iv
schmidt (espe cially f or her teaching mentorship and job market prep), Robin Jeshion and
Gabriel Uzquiano (f or their guidanc e as DGS), and Jim V an Clev e (f or his generous ac c om-
modations as I re c o v ere d from b ack surgery during m y first y ear).
And then there’s m y dissertation c ommitte e: M ark Schroe der (c o-chair), Jon Quong (c o-
chair), Jake Ross, Jake Nebel, and Greg Keating. What a bunch. I ha v e Greg to thank f or
introducing me to the Philosop h y of La w during m y first y ear and f or helping me na vigate
the interse ction betw e en m y p hilosop hic al research and the la w . I am thankful to Jake Nebel
f or alw a ys being a v ailab le, and f or man y rich discussions. I am thankful to Jake Ross f or
being a c ountere x amp le machine from whom no argument or princip le is saf e. I’ll miss our
spontane ous p hone c alls, and trying to c on vinc e Jake that he’s got the wrong intuitions.
As f or M ark and Jon: whatev er I c ould sa y ne xt w ould f ail to do them justic e and w ould
f ail to ade quately e xpress m y gratitude. They are mentor-advisers greater than which none
c an be c onc eiv e d. M ark has helpe d me na vigate ev ery step of the journey from area e x am to
qualifying e x am to dissertation to job market. He has read dozens of drafts. He has w alke d
dozens of c aff einate d laps around c ampus with me. He has talke d me through dozens of
prof essional and personal anxieties. I ha v e learne d much as a p hilosop her from M ark. He is
e x c eptional at helping one to se e the “big picture” and important imp lic ations of one’s views,
and of teasing out unnotic e d interc onne ctions. He is a w orld-class practitioner of the dark
art of writing a narrativ ely-c ompelling p hilosop h y p aper, and I’v e learne d much from him in
v
this respe ct. He is also an e x emp lary and c ompelling model of some one who c ares about the
health of the dep artment and the discip line. But as grateful as I am f or all of the p hilosop hic al
and prof essional training I’v e re c eiv e d from M ark, I’m ev en more grateful f or his ca re . M ark is
de ep ly in v este d in the w ell-being of each of his (man y!) advise es, in w a ys that go w ell bey ond
their prof essional liv es. He fre quently opens his home to his advise es, che cks in often, and
makes himself a v ailab le on short notic e no matter ho w bus y . I’ll miss him dearly , and I’ll miss
his lo v ely f amily .
Jon w as the first person I T A ’ d f or at USC, and I nev er loo ke d b ack. I knew imme diately I
ne e de d to w ork with him, and I’v e glomme d on to Jon like a p arasite f or the last fiv e y ears. I’v e
leache d so much of his time that I don’t kno w ho w he’s still standing. Jon has prob ab ly read
— I don’t e x aggerate — fifty drafts in some f orm or other. And someho w his c omments ha v e
nev er f aile d to be e xtensiv e, detaile d, brilliant, a nd f ocusing. M ari c an attest to the e x citement
with which I go into and out of offic e hours with Jon: spending time on the p hilosop hic al
danc efloor with Jon is a w orkout and a jo y . Jon has also be en a tremendous teaching model
and mentor, and I aspire to his clear, engaging, and kind teaching style, and his appro achabil-
ity in offic e hours. As a kid, I w ante d to be all sorts of diff erent athletes or superheroes when
I grew up; no w I just w ant to gro w up to be Jonathan Quong.
And then there’s some one who I c ould nev er ev en aspire to be: Natalie Schaad. Natalie
is the Philosop h y Dep artment’s program administrator, though that’s the most inade quate
vi
job title f or some one who does what Natalie does — which is, everyt hi ng . She has masterfully
na vigate d our dep artment through treacherous administrativ e w aters that w ould ha v e sunk
an y other ship . And she’s done this o v er and o v er and o v er again. (“Can w e fix it?” “Y es,
Nata li e c an!”) Someho w , though, Natalie has done all this while also being endlessly a v ailab le
to graduate students. While Natalie w as still with us in person, her door w as alw a ys open. I
droppe d in c onstantly , and chatte d her ear off when she no doubt had much more important
things to f ocus on. But it w as hard to lea v e her offic e! — be c ause talking to Natalie w as alw a ys
such a jo y , and be c ause she w as an endless f ount of kindness and sage advic e (whether about
administrativ e matters or personal ones). I’ d ne e d to write another dissertation to ade quately
ac c ount f or all the things f or which I am grateful to Natalie.
There are man y to ackno wle dge outside of ac ademia, without whose support and friend-
ship I’ d ha v e nev er surviv e d this chapter. The v ery first I should mention is the Nimmo f amily ,
who house d me b ack in m y undergrad da ys, and who ha v e c are d f or m y f amily throughout
our se c ond stint in Calif ornia. Jessic a and Da v e Nimmo ha v e treate d me like a son, M ari like
a daughter, and m y children like their grandkids. They ha v e spoile d us with their lo v e, and I
c annot thank them enough. I am also thankful to Eric Nimmo , f or f orcing me to w ake up
early , putting up with m y dogge d p ac e on our man y Turnbull Can y on runs, and f or man y
fun, o xygen-depriv e d arguments.
I am also thankful to Tim and Sarah Buchanan. They ha v e be en e x c eptionally lo y al and
vii
c onstant friends, who ha v e steppe d up f or m y f amily time and again: making meals when w e
w ere sick, w atching the kids on short notic e, or bringing o v er bo ard games and drinks when
that w as what w e ne e de d most. I o w e them more than I c an hope to rep a y .
M y thanks also to the man y w onderful f o lks at Anglic an Church of the Epip han y , who
ha v e be en an e xtende d f amily to us the last six y ears, c aring f or us in more w a ys than I c an tell
— but espe cially to Fr. Greg P eters and Christina P eters, who ha v e supporte d M ari and I in
v arious w a ys f or almost f ourte en y ears!
I’m also thankful f or all m y ultimate frisbe e mates in La Mirada and Fullerton, but es-
pe cially: Da vid F onzi, Da vid Shelton, Alissa “Cricket” Hiramine, Nathaniel Ankrum, Phil
Hooton, and Zuzu P asch. Pickup and league frisbe e kept me sane through it all. Thanks f or
the great fun, the a w esome e x ercise, and the ev en better c amaraderie.
M y in-la ws, Kath y and Stev e P ac e, ha v e be en a steady sourc e of support during their man y
visits from the P acific North w est. I am espe cially grateful f or ho w intensely they lo v e m y
children, but I am also grateful f or their c omp an y , their respe ct, and their ha ving alw a ys w el-
c ome d me into their f amily .
M y sister, Cassie, has alw a ys be en and will alw a ys be one of m y closest friends. Despite her
ha ving f our kids under the age of eight, and despite m y brotherly tomf oo lery , she c ontinues
to be one of m y strongest and most regular emotional supports and one of m y v ery f a v orite
pe op le to talk with. I am thankful f or her lo v e.
viii
I am also thankful to (and f or) each of m y children: M ap le, Miah, and W endell. I’m thank-
ful f or their beautiful and wildly diff erent personalities, their boundless aff e ction, and the
purpose and f ocus they’v e besto w e d on m y lif e. They’v e made m y time in graduate schoo l
v ery bus y , but inc re dib ly fun. F or that I am thankful. I’m also thankful f or the rare oc c asions
on which they w ere quiet while I w as trying to write.
I c an’t ade quately thank m y dad, Stev en Clark. I’v e put him through a lot o v er the y ears,
and w e’v e be en through a lot together. He has be en a steadf ast support through it all. He’s
driv en mo ving trucks ac ross the c ountry , sat with me f or da ys in a hospital, driv en eight hours
through sno w to deliv er me heating oil, w atche d m y dog f or a semester, allo w e d me to fill
his too l she d with f arm animals bef ore I ac cidentally burne d it to the ground — and that’s
someho w the v ery least of it. I’m so proud to c all him m y dad.
Finally , there’s M ari, m y wif e, to whom m y gratitude takes me w ell bey ond the e xpressiv e
limits of the English language. I definitely w ould not be here without her support and en-
c ouragement. M ari’s productivity is humb ling, her p atienc e is saintly , and her c onfidenc e in
me is wildly undeserv e d (m y kids are not so easily f oo le d). I f e el v ery b lesse d, e x c ept f or the
p art where M ari is so much better than me at almost ev erything and makes me loo k b ad.
I ha v e be en giv en so much b y so man y . Thank y ou all.
ix
T ab le of Contents
0 Abstract 1
1 Chapter One: The Demands of Necessity 5
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Existing Ac c ounts of Ne c essity are Unsatisfying . . . . . . . . . . . . . . . 8
1.3 Liability , Duty , & Proporti onality . . . . . . . . . . . . . . . . . . . . . . 18
1.4 The Demands of Ne c essity . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.5 Ne c essity as Responsibility -Sensitiv e . . . . . . . . . . . . . . . . . . . . 32
1.6 Multip le Attackers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
vii
2 Chapter Two: Refusing Protection 40
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.2 T w o Ingre dients f or P ermissib le Def ense . . . . . . . . . . . . . . . . . . 43
2.3 The Ex clusion Ac c ount . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.4 The Liability Contraction A c c ount . . . . . . . . . . . . . . . . . . . . . 55
2.5 F eatures and Imp lic ations . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.6 State Officials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3 Chapter Three: Defense Without Threat 77
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.2 The Challenge of M erely App arent Attackers . . . . . . . . . . . . . . . . 81
3.3 Culp ability , Responsibil ity , and Signaling . . . . . . . . . . . . . . . . . . 84
3.4 A New Answ er . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.5 Adv antages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4 Chapter Four: The Price of Duty 110
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.2 The Princip al-Agent Relati on . . . . . . . . . . . . . . . . . . . . . . . . 113
viii
4.3 The Benefiary P a ys Ac c ount . . . . . . . . . . . . . . . . . . . . . . . . . 120
4.4 The Pric e of Duty Ac c ount . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.5 Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
References 143
ix
0
Abstract
Respondi ng to Ha rm is a c o lle ction of essa ys c onc erning the ethics of harm. The first thre e
chapters c onc ern the subje ct of def ensiv e or prev entativ e harm — that is, harm that is im-
pose d on some one to prev ent harm to some one else. The final chapter c onc erns c ompen-
satory morality; it c onc erns the nature of our duties to p a y c ompensation f or harms that
ha v e already oc curre d. All f our chapters e xp lic ate, in diff erent w a ys, the nature and moral
1
signific anc e of our duties to assume c osts from others.
Chap ter 1: “The Demands of N e c essity” . It is widely held that def ensiv e harm must be
both ne c essary and proportionate. Y ou shouldn’t kill some one in self-def ense who is only
trying to pinch y ou on the arm; that w ould be disproporti onate . But ev en when an attacker
poses a lethal threat, y ou shouldn’t kill them in self-def ense if y ou c an just as w ell stop their
attack with a slap on the wrist; killing them w ould then be u n nec essa ry . In chapter one I
dev elop a no v el e xp lanation, pre cisific ation, and unific ation of the proportionality and ne-
c essity c onstraints. What se em to be tw o independent c onstraints, I argue, are not. Unjust
attackers ha v e a duty to a v ert their o wn attack. This duty c ould re quire them to bear up to
a c ertain amount of c ost to a v ert their attack w ere they ab le. And this in turn determines
what others ma y do to them: an attacker ma y only be made to bear as much c ost to a v ert
their attack as they c ould be re quire d to take upon themselv es. The proportionality and ne-
c essity c onstraints, I argue, both e xpress this fundamental princip le, only at diff erent lev els
of generality .
Chap ter 2: “Refusing Pr o te ction” . The se c ond chapter takes up a diff erent question
about the limits of def ensiv e harm. Just as w e c an often use f orc e to def end ourselv es, so too
c an w e often use f orc e to def end others. But suppose some one doesn’t w ant our prote ction.
A c ommitte d p acifist asks y ou not to use vio lenc e to prote ct her. A po w erful aggressor na-
tion in v ades a w eaker neighbor, but the w eaker nation v otes against w elc oming the military
interv ention of the Unite d States. What then? I argue that such refusal sometimes makes it
wrong to interv ene. This po w er is an upshot of a more fundamental po w er w e each ha v e to
alter the “ enf orc eability” of our rights. It’s f amiliar that w e ha v e po w ers to transf er or w aiv e
rights. F or the same reasons that w e ha v e these po w ers o v er our rights, I argue, w e ha v e the
2
po w er to change whether a right ma y be enf orc e d, and b y whom. One w a y in which this
po w er c an be e x ercise d is b y the refusal of prote ction: b y refusing m y prote ction y ou remo v e
me from the class of persons with respe ct to whom y our attacker is liab le, thereb y making it
so that you r attacker is not liab le to def ensiv e harm at m y hands. This chapter def ends the
idea that w e ha v e such po w ers o v er our rights, and e xp lores the imp lic ations of this idea f or
un w ante d military and po lic e interv entions.
Chap ter 3: “Def ense W ithout Thr eat” . This chapter c ombines and dev elops some of
the ideas from the first tw o chapters to answ er an important challenge f or the ethics of de-
f ense. Consider a p aradigmatic c ase of self-def ense. A villain attempts to murder y ou, and
y ou’ll be kille d unless y ou kill him first. The villain is clearly lia b le to be kille d b y attacking
y ou — that is, he lacks his normal right that y ou not kill him, which largely e xp lains wh y y ou
are permitte d to kill him.
The villain’s liability sure se ems to ha v e something to do with the f act that he wou ld ha rm
someone if he isn’t ha rmed hi m fi rst . That se ems, at first glanc e, at least a ne c essary c ondi-
tion f or liability . But there’s a prob lem. Suppose a hitman attempts to kill y ou, but unbe-
kno wnst to y ou their gun is jamme d. Or suppose a b ank robber ”b luffs ” that he will kill y ou
if y ou don’t open the v ault. Y ou don’t ne e d to harm the hitman or the b ank robber to pre-
v ent y ourself from being harme d. But it sure se ems they are nonetheless liab le to be harme d,
una w are as y ou that they pose no actual threat. Our the ory of liability ne e ds to speak to such
c ases. I c all this the Challenge of M erely App arent Attackers.
Chapter thre e presents an unorthodo x answ er to this challenge. I dev elop and def end a
the ory of liability ac c ording to which merely app arent attackers are not strictly liab le to be
harme d. Their rights against harm persist, and those rights c ontinue to giv e us reason not
3
to harm them. Nonetheless, characters like the hitman and the b ank robber ha v e a duty to
assume the c osts of the def ensiv e actions that are taken against them, and this duty e xp lains
wh y they lack man y of the usual perks or upshots of their rights. They are what I c all quasi-
lia b le . Their moral status is neither that of an innoc ent person nor that of an attacker who
poses a genuine threat of harm.
Chap ter 4: “The Pric e of Duty” . State officials regularly impose harm on the citizens
they are suppose d to serv e, some of it wrongful. Who should bear the burden of p a ying
c ompensation f or these wrongs? Should it be the agents themselv es, or should the burden
be spread ac ross the citizenry in the f orm of tax ation? In chapter f our I def end a the ory of
li mited offi cia l i m mu nity , ac c ording to which the citizenry should assume the c osts of c ertain
official wrongs. I argue that limite d official immunity is an upshot of a general princip le of
distributiv e justic e, ac c ording to which those who are morally re quire d to p articip ate in some
proje ct are re quire d to share c ertain c osts of that proje ct. Understanding official immunity
as deriv ativ e of this princip le helps us to identify , not just the grounds, but also the li mits of
official immunity: it helps us to sort betw e en the kinds of wrongs f or which the burdens of
official wrongs should f all on the individual agents themselv es and the kinds of wrongs f or
which those burdens should be spread ac ross the citizenry .
4
A n act t hat proc eeds from a good i ntenti on ca n be rendered
i l li cit if it is not proporti onate to t he end. A nd so if, i n order
to defend his own life, someone uses more vi o lenc e t ha n is
nec essa ry, t he act wi l l be i l li cit.
- Thomas Aquinas
1
Chapter One: The Demands of Ne c essity
1.1 Introduction
M oral p hilosop hers ha v e long re c ognize d a proporti ona lity c onstraint on def ensiv e harm.
1
Comp are:
F ast a nd Fu ri ous. Re ckless loses c ontro l of his c ar in an illegal ro ad rac e, endan-
1
The quote from Aquinas c omes from Su m ma Theo logi ca , Pt.II-II, Q64, art. 7.
5
gering an innoc ent pe destrian. P e destrian c an let the c ar hit her, in which c ase
she will be permanently p aralyze d from the ne ck do wn. Or she c an re dire ct the
c ar a w a y from he r, in which c ase Re ckless will be permanently p aralyze d.
F ast a nd Less Fu ri ous . As bef ore, e x c ept the threat to P e destrian is much smaller.
If P e destrian lets the c ar hit her, she will only suff er a small bruise.
P e destrian is permitte d to re dire ct the c ar in the first c ase but not the se c ond. The relev ant
diff erenc e, it se ems, has to do with the sev erity of the threat a v erte d. In the first c ase the
def ensiv e harm se ems proportionate to the threat a v erte d (p aralysis v . p aralysis). In the se c ond
c ase the harm se ems disproportionate; it se ems “too great” relativ e to the sev erity of the threat
a v erte d (p aralysis v . a small bruise). As this p air of c ases illustrates, def ensiv e harm should be
(in some sense) proportionate to the threat a v erte d. This idea is widely ac c epte d b y the orists
and la ype op le alike, and embe dde d in the la w .
2
The proportionality c onstraint is standardly distinguishe d from another widely-ac c epte d
c onstraint — the nec essity c onstraint.
3
Consider:
Easy Defense . As in F ast and Furious, e x c ept that P e destrian has tw o w a ys to
re dire ct the c ar. If she re dire cts the c ar to her right, Re ckless w ill be p aralyze d.
If she re dire ct the c ar to her left, Re ckless will suff er a spraine d toe. P e destrian
chooses t o p aralyze Re ckless.
2
Americ an La w Institute, Mode l Pena l Code (Philadelp hia: The Institute, 1985), §3.04-05.
3
What I’m here c alling the ‘proportionality c onstraint’ is sometimes ref erre d to in the literature as the ‘nar-
ro w proportionality c onstraint’, to be distinguishe d from the ‘wide proportionality c onstraint’. Despite the f act
that both princip les impose re quirements of proportionality , they are importantly diff erent from one another.
The narro w proportionality c onstraint f ocuses on the attacker; it re quires that the def ensiv e harm impose d
on the attacker be proportionate to his attack. The wide proportionality c onstraint, b y c ontrast, f ocuses not
(just) on the attacker, but on innoc ent p arties; it re quires that aggregate “ c o llateral damage” be proportionate to
the unjust threat that is a v erte d. Se e Jeff M cM ahan, Ki l li ng i n W a r (Oxf ord: Oxf ord Univ ersity Press, 2009),
22-24.
6
P e destrian acts impermissib ly . The prob lem, ho w ev er, isn’t that it is disproportionate to
p aralyze an attacker to def end against p aralysis (w e’v e se en abo v e that it isn’t). Rather, P e des-
trian’s action is impermissib le, it se ems, be c ause she had another def ensiv e option a v ailab le to
her that w ould ha v e impose d much less c ost on Re ckless without imposing much (or in this
c ase, an y) e xtra c ost on P e destrian. P e destrian didn’t “ne e d ” to p aralyze Re ckless to def end
herself.
4
Despite the longstanding and wide re c ognition of the proportionality and ne c essity c on-
straints, there has be en little discussion of their relationship to one another. This, I think, is
be c ause the orists ha v e generally taken the tw o c onstraints to be e xp laine d b y diff erent moral
c onsiderations.
5
In what f o llo ws I push b ack against this c ommon appro ach. I dev elop and
motiv ate a the ory ac c ording to which “proportionality” and “ne c essity” are tw o f ac es of a
c ommon moral demand. While there is a diff erenc e in sc ope betw e en the tw o c onstraints,
disproportionate harm and unne c essary harm wrong an attacker in the same w a y: b y making
him to bear more c ost in servic e of some def ensiv e go al than he w ould be re quire d to take
upon himself in servic e of that go al. At the end of the da y , the tw o c onstraints e xpress this
same fundamental princip le, only at diff erent lev els of generality .
The p aper proc e e ds as f o llo ws. §1.2 c onsiders the tw o most prominent and dev elope d ac-
c ounts of the ne c essity c onstraint. I e xp lain wh y these ac c ounts are unsatisfying. §1.3 c onsid-
ers the nature of the proportionality c onstraint. I argue that this c onstraint is best e xp laine d
b y c ertain f eatures of an attacker’s duties, bef ore sho wing in §1.4 ho w this e xp lanation c an be
4
F or a discussion of the ne c essity c onstraint in diff erent legal s ystems, se e Ge orge Fletcher, Ret hi n ki ng Cri m-
i na l Law (Oxf ord: Oxf ord Univ ersity Press, 2000), 855-875. Se e also the Mode l Pena l Code , §3.04.
5
Se e, f or e x amp le, Da vid Rodin, W a r a nd Se lf-Defense (Oxf ord: Oxf ord Univ ersity Press, 2002), 40-41,
Helen Fro w e, Defensive Ki l li ng (Oxf ord: Oxf ord Univ ersity Press, 2014), 118-119, and Jonathan Quong, The
Mor a lity of Defensive Forc e (Oxf ord: Oxf ord Univ ersity Press, 2020), 129-132.
7
generalize d to ac c ount f or the ne c essity c onstraint. The ne xt tw o se ctions c onsider what f o l-
lo ws from this ac c ount of ne c essity: §1.5 c onsiders what f o llo ws f or the relationship betw e en
responsibility and ne c essity , and §1.6 c onsiders what f o llo ws f or c ases in v o lving multip le at-
tackers. §1.7 c oncludes.
1.2 Existing Accounts of Nece ssity are Unsatisfying
1.2.1 The Discount Account
There is a c onstraint against “unne c essary” def ensiv e harm. But what is it f or harm to be
unne c essary in the relev ant sense? T aking talk of ‘ne c essity’ at f ac e v alue, here’s a natural
answ er: def ensiv e harm is unne c essary just in c ase it is possi b le to get as much def ensiv e gain
without imposing as much harm on the attacker or other innoc ent persons.
The orists such as Seth Lazar and Jeff M cM ahan ha v e observ e d, ho w ev er, that this natural
gloss c an’t be the right w a y to think about the ne c essity c onstraint.
6
T o se e wh y , c onsider a
v ariation on the E as y Def ense c ase:
Easyis h Defense . As in E as y Def ense, e x c ept that P e destrian will suff er a mild
spraine d toe of her o wn if she chooses to sprain Re ckless ’s toe rather than p ar-
alyze him. P e destrian chooses to p aralyze Re ckless.
In this c ase p aralyzing Re ckless off ers more def ensiv e gain than an y other option; it is not
possib le f or P e destrian to just as w ell prote ct herself b y other means. Thus, on the naiv e
reading of “ne c essity” desc ribe d abo v e, P e destrian does not vio late the ne c essity c onstraint.
6
Seth Lazar, “Ne c essity in Self-Def ense and W ar,” Phi losop hy a nd Pu b li c Aff a i rs 40 (2012): 2-23; Jeff M cM a-
han, “The Limits of Self-Def ense,” in The Et hi cs of Se lf-Defense , e ds. Christian Coons and Michael W eber
(Oxf ord: Oxf ord Univ ersity Press, 2016): 185-189.
8
M an y the orists find this an imp lausib le result.
7
W e should treat like c ases alike if w e c an,
and intuitiv ely the sort of wrong disp la y e d i n E as y Def ense is the same sort of wrong disp la y e d
in E as y is h Def ense. Roughly put, P e destrian acts wrongfully in both c ases in that she had
another def ensiv e option a v ailab le to her that w ould ha v e impose d mu c h less c ost on Re ckless
while imposing little or no e xtra c ost on P e destrian.
The question is ho w to make this rough diagnosis pre cise.
Lazar and M cM ahan ha v e each def ende d a pre cisific ation on which the ne c essity c onstraint
is a spe cial instanc e of a general moral re quirement to minimize harm. Sa ys M cM ahan:
The morally best and theref ore ne c essary def ensiv e option is the one that, in
c omp arison with other options … has the morally best trade-off betw e en the
harm prev ente d and the harm c ause d … taking into ac c ount the prob ability of
suc c ess, the e xpe cte d c ost to the agent, the e xpe cte d harm to others as a side
eff e ct, and the e xpe cte d harm to the threatener.
8
Of c ourse, a def ender is not re quire d to minimize harm fu l l-stop . Suppose a def ender c an
ke ep an attacker from breaking her leg only b y breaking his leg a nd one of his toes. By so
def ending herself she w ould f ail to minimize total harm. But she p lainly does not vio late the
ne c essity c onstraint. Sa ys Lazar: this is be c ause in c ases of def ensiv e harm, the interests of
7
F or an e x c eption, se e Kaila Draper, “Ne c essity and Proportionality in Def ense,” in The Et hi cs of Se lf-
Defense , e ds. Christian Coons and Michael W eber (Oxf ord: Oxf ord Univ ersity Press, 2016): 180-181. Draper
argues that whereas the prob lem with c ases like E as y Def ense is that the def ensiv e harm is unne c essary , the
prob lem with c ases like E as yish Def ense is that the def ensiv e harm is disproportionate.
8
Jeff M cM ahan, “Ne c essity and Proportionality in M orality and La w ,” in Nec essity a nd Proporti ona lity i n
Internati ona l Peac e a nd Secu rity Law , e ds. Claus Kres and Robert La wless (New Y ork: Oxf ord Univ ersity
Press, 2021): 23; M cM ahan, “Limits of Self-Def ense,” 187-188.
9
the attacker are disc ou nted relativ e to the interests of innoc ent persons.
9
What the ne c essity
c onstraint re quires is not that def enders minimize harm, but that they minimize “morally-
w eighte d ” harm.
Let’s c all the ac c ount def ende d b y Lazar and M cM ahan:
Discount . The imposition of def ensiv e harm on an attacker is unne c essary
just in c ase there are other def ensiv e options that impose less e xpe cte d, aggre-
gate, mora lly-w eighte d harm.
10
So long as the disc ounte d interests of the attacker still c ount f or something, the Discount
the ory e xp lains wh y it is unne c essary to p aralyze Re ckless is E as y Def ense. P e destrian c an
p aralyze Re ckless or merely sprain his toe. Sinc e both options do e qually w ell at prev enting
harm to persons other than Re ckless, P e destrian w ould f ail to minimize morally-w eighte d
harm b y p aralyzing Re ckless.
Likewise, the ac c ount has the resourc es to e xp lain wh y P e destrian ma y not p aralyze Re ck-
less in E as y is h Def ense (where refraining from p aralyzing Re ckless will c ost P e destrian a spraine d
toe). So long as the interests of attackers are not disc ounte d too greatly , Re ckless ’s disc ounte d
harm of p aralysis will still be greater than the aggregate of a disc ounte d spraine d toe f or Re ck-
less p lus a non-disc ounte d spraine d toe f or P e destrian.
Discount has its virtues. It c aptures our intuitiv e v erdicts in c ases like E as y Def ense and
E as yish Def ense. And it deliv ers a gain in the oretic al unity as it ties together the ne c essity
9
Lazar, “Ne c essity ,” 6. M cM ahan is no where e xp licit about whether and ho w the interests of attackers
are disc ounte d. But his judgments about p articular c ases together with his claims about the ne c essity c onstraint
clearly c ommit him to the idea that the interests of (responsib le) attackers are disc ounte d relativ e to the interests
of innoc ent persons. Se e also Da vid Rodin, “Justifying Harm,” Et hi cs 122 (2011): 99.
10
M cM ahan, “Limits of Self-Def ense,” 187-188; Lazar, “Ne c essity ,” 13. F or heuristic purposes I am ignoring
some of the c omp le xities that c onc ern prob ability trade-offs, sinc e all of the c ases that appear in this p aper in v o lv e
e quiprob ab le def ensiv e options.
10
c onstraint and the so-c alle d “wide” proportionality c onstraint, grounding both c onstraints
in the p lausib le idea that w e should minimize morally-w eighte d, aggregate harm.
11
But despite these virtues, I think w e should reje ct Discount . The the ory suff ers tw o
serious shortc omings. The first is that it off ers imp lausib le v erdicts in c ertain c ases. Consider:
Nec essa ry Ki l li ng . Re ckless loses c ontro l of his c ar, which will permanently p ar-
alyze P e destrian from the w aist do wn unless she re dire cts the c ar, killing Re ck-
less.
Un nec essa ry Ki l li ng . As abo v e, e x c ept that P e destrian c an def end herself either
b y killing Re ckless or b y merely spraining one of his fingers.
Un nec essa ry Ki l li ng (Bysta nders) . Again, P e destrian c an def end herself either b y
killing Re ckless or b y merely spraining one of his fingers. Killing Re ckless in-
v o lv es re dire cting his c ar into the p ath of an unrelate d a v alanche. The a v alanche
will kill Re ckless, but his c ar will b lock the a v alanche from killing tw o innoc ent
b ystanders.
In the first c ase — Ne c essary Killing — P e destrian is p lainly permitte d to kill Re ckless.
Discount is c onsistent with this result so long as Re ckless ’s interests are disc ounte d. If his
interests are disc ounte d, then there is no f ailure to minimize total, morally-w eighte d harm b y
killing him.
In the se c ond c ase — Unne c essary Killing — P e destrian is p lainly not permitte d to kill
Re ckless. Doing so w ould be a p a radigm vio lation of the ne c essity c onstraint. And Dis-
count off ers an e xp lanation wh y: killing Re ckless w ould be unne c essary be c ause it f ails to
minimize total, morally-w eighte d harm (spraining the finger does better in this respe ct).
11
Se e f ootnote #3 on the distinction betw e en “wide” and “nar ro w” proportionality .
11
But if it is unne c essary to kill Re ckless in the se c ond c ase, then it is surely also unne c essary
to kill Re ckless in the third c ase. Discount pre dicts otherwise, ho w ev er, sinc e this third
option does minimize total morally-w eighte d harm. This isn’t to sa y that Discount is in-
c omp atib le with the f act that it is wrong to kill Re ckless in this c ase: perhaps there is some
other moral c onstraint that f orbids P e destrian from using Re ckless to b lock the a v alanche.
But it is to sa y that Discount is inc omp atib le with the f act that killing Re ckless is u n nec-
essa ry — which it surely is, and which it surely is for t he sa me reasons that killing Re ckless is
unne c essary in the c ase where no b ystanders are present.
This brings me to m y more serious w orry about Discount . It’s not just that it deliv ers
strange results in p articular c ases, but that it misi dentifi es the wrongfulness of unne c essary
harm. P aradigm c ases of unne c essary def ensiv e harm are wrongful (at least in p art) be c ause
they wrong the person harme d. An ac c ount of the ne c essity c onstraint should identify this
distinctiv e, dire cte d wrong. But Discount does not identify an y such dire cte d wrong. On
the c ontrary , it characterizes the wrongfulness of unne c essary harm as i mpersona l in nature.
12
Let’s u np ack this w orry . Start with m y claim that unne c essary harm in v o lv es a di rected
wrong. W e c an test this claim. T o test whether the imposition of some def ensiv e harm H
wrongs its target w e ask whether the def ender, b y attempting to impose H, makes himself
liab le to c ou nter-defensive harm from a third-p arty .
13
If the def ender does make himself liab le
to such c ounter-def ensiv e f orc e, this is good evidenc e that the imposition of H w ould wrong
its target.
14
This is be c ause a person does not make themselv es liab le to def ensiv e harm simp ly
12
C.f. Quong, “M orality of Def ensiv e F orc e,” 130.
13
As I use the term ‘liab le’ throughout the p aper: f or X to be liab le to the imposition of harm H b y Y is f or
X to lack a right against Y’s imposition of H ev en if X does not c onsent to suff er H.
14
In some c ases, a def ender who uses unne c essary f orc e ma y be liab le ev en to c ounter-def ense b y her attac ker .
But it is easier f or a def ender to make herself liab le to c ounter-def ense b y a third p arty than b y her attacker; this
is wh y I f ocus on the third-p arty c ase abo v e.
12
b y posing a threat of harm; one makes themselv es liab le to def ensiv e harm only b y posing a
threat of rig hts-i nfri ngi ng harm. (Otherwise, ev ery def ender w ould make themselv es liab le
to c ounter-def ensiv e harm when they fight b ack against their attacker.)
Let’s app ly this test in the c ase of def ensiv e harm that is p lainly unne c essary . Consider a
v ariant of E as y Def ense.
15
P e destrian c an p aralyze Re ckless or sprain his toe. She chooses
to p aralyze him. But this time there is a b ystander, Blocker, who c an interv ene to prote ct
Re ckless from P e destrian’s unne c essary harm. If Blocker interv enes, Re ckless and P e destrian
will each suff er just a spraine d toe.
The question is whether Blocker w ould wrong P e destrian b y interv ening. I find it intu-
itiv ely ob vious that she does not.
16
M ore o v er, none of the usual evidential markers of a wrong
are present: f or e x amp le, P e destrian ma y not fight b ack against Blocker’s interv ention and
P e destrian is not o w e d c ompensation if Blocker’s interv ention suc c e e ds. There is both dire ct
and indire ct intuitiv e support f or the judgment that Blocker’s interv ention does not wrong
P e destrian.
But if Blocker’s interv ention does not wrong P e destrian, then it f o llo ws that P e destrian
made herself lia b le to Blocker’s interv ention. As w e note d abo v e, P e destrian’s liability is
strong evidenc e that P e destrian’s attempt to impose u nne c essary harm is not just imperson-
ally wrongful, but that it threatene d Re ckless ’s rig hts .
17
T o generalize: unne c essary harm is
wrongful (at least in large p art) be c ause it wrongs the person harme d.
15
A v ersion of the f o llo wing argument appears in Dr aper, “Ne c essity and Proportionality , 174-175.
16
I find this v erdict ev en more ob vious if w e tw eak the c ase such that Re ckless poses a wrongful threat, but
not one f or which he is culp ab le. P erhaps he reasonab ly , but mistakenly , believ es he is racing on a designate d
racing c ourse.
17
P erhaps surprisingly , M cM ahan himself has def ende d the claim that unne c essary harm wrongs its target.
Se e M cM ahan, “Limits of Self-Def ense,” 195-197. There are some who obje ct to the claim, ho w ev er. Fro w e,
Defensive Ki l li ng , 88-119 is a notab le e x amp le. F or a response to (some of) Fro w e’s arguments, se e M cM ahan,
“Limits of Self-Def ense,” 195-206.
13
Discount , ho w ev er, goes no w a ys to w ards identifying this wrong. Discount appeals to
an impersonal duty to minimize total (morally-w eighte d) harm, but the f ailure to minimize
harm is no more a f ailure to respe ct the rights of one’s attacker than the rights of an y other
person. Discount thus does not help us to identify the distinctiv e w a y in which unne c essary
harm wrongs the person(s) harme d — which ought to be a c entral desideratum on an ac c ount
of the ne c essity c onstraint.
18
1.2.2 The Rescue Account
There is one ac c ount in the literature that makes it a point to e xp lain the wrongfulness of
unne c essary harm b y appeal to a dire cte d wrong: the “Rescue Ac c ount”, first presente d b y
Jo anna Firth and Jonathan Quong, and later dev elope d in greater detail b y Quong.
19
On
this ac c ount, the ne c essity c onstraint is grounde d in “Good Samaritan” duties of rescue that
def enders o w e to attackers.
The idea is this. All persons ha v e a right to eas y rescue — a right to be rescue d from serious
harm when the c ost to the rescuer (and b ystanders) is small. These rights are so b asic to one’s
moral status as a human being that they are not f orf eite d ev en b y unjust attackers; attackers
maintain these rights ev en in the midst of their attack.
20
One w a y to rescue a person is to
refrain from imposing harm on them that y ou are otherwise at liberty to impose. Thus, if a
def ender c an refrain from imposing sev ere harm on an attacker at small c ost to herself and
other innoc ent p arties, then the def ender has a duty to do so . Imposing the sev ere harm w ould
18
None of this is to sa y that c onsiderations of harm minimization p la y no important function in the ethics
of de f ensiv e harm. They surely do . I’m just arguing that the ne c essity c onstraint is not itself a re quirement of
harm minimization.
19
Jo anna M ary Firth & Jonathan Quong, “Ne c essity , M oral Liability , and Def ensiv e Harm,” Law a nd Phi-
losop hy 31 (2012): 673-701; Quong, Mor a lity of Defensive Forc e , 124-149.
20
Quong, Mor a lity of Defensive Forc e , 131.
14
vio late the unjust attacker’s right to eas y rescue. This, claims Quong, is wh y it is wrong f or
P e destrian to p aralyze Re ckless regardless of whether this will re quire Def ender to suff er a
spraine d toe.
21
Rescue. The imposition of def ensiv e harm is unne c essary just in c ase it w ould
infringe t he attacker’s right to eas y rescue.
The ac c ount is pref erab le to Discount in the respe ct that it purports to e xp lain wh y
unne c essary harm wrongs the attacker. But Rescue , I think, appeals to the wrong sort of
dire cte d duty to get this result. The appeal to duties of eas y rescue yields un w elc ome results.
One challenge f or the ac c ount c onc erns the f act that — as Quong himself ackno wle dges
— duties of rescue se em to arise only when the harm the rescuer c an a v ert is quite serious.
22
Duties of rescue do not arise whenev er there is simp ly a large enough r ati o betw e en the c osts
to be a v erte d and the c osts of rescue. A stranger has a right that y ou rescue them from seri ous
harm when the c ost to y ou is small. But a stranger does not ha v e a right that y ou rescue them
from sma l l harm when the c ost to y ou is v ery small. Suppose I trip on m y w a y to le cture and
drop m y stack of p apers. I instinctiv ely reach do wn to c atch it, which will c ause me a number
of v ery minor p apercuts. Y ou se e that y ou c an interv ene: y ou c an c atch the p apers bef ore I
do , in which c ase y ou will suff er just one minor p apercut. Here it is p lausib le that y ou s hou ld
interv ene. But most of us don’t think I ha v e a rig ht that y ou do this f or me; y ou don’t owe it
to me to suff er the p apercut. This is evidenc e d b y the f act that I don’t ha v e the standing to
demand that y ou rescue me nor the standing to c riticize y ou if y ou choose not to rescue me.
Things are v ery diff erent when the stakes are raise d: if y ou c an rescue me from dro wning at
21
Ibid., 132.
22
Ibid., 132.
15
small c ost to y ourself, I do ha v e the standing to demand y our rescue and to c riticize y ou if
y ou choose to let me dro wn.
Here is wh y this makes troub le f or Rescue . Be c ause duties of rescue only arise when
the threat is serious, Rescue f ails to c ount small, but entirely pointless def ensiv e harms as
unne c essary . Suppose I lose c ontro l of m y c ar and it sw erv es to w ards y ou. Y ou c an e qually
w ell def end y ourself b y (i) breaking m y leg or b y (ii) breaking m y leg a nd spraining m y big toe.
The more harmful option is p lainly impermissib le and unne c essary . After all, spraining m y
toe w ould be entirely gratuitous. But if w e do not ha v e duties to rescue strangers from small
harms, then Rescue does not deliv er this result. The harm of a spraine d toe is too small
to impose on y ou a duty of rescue. Rescue thus off ers no help e xp laining what’s wrong
with small, but entirely gratuitous def ensiv e harm. It tells us that such harms are not, in f act,
unne c essary .
Quong ackno wle dges this imp lic ation, and is willing to ac c ept it.
23
But I don’t think w e
should be so willing. This result is imp lausib le in its o wn right. But it loo ks ev en more im-
p lausib le when w e c onsider a further imp lic ations of this result. T o illustrate the prob lem,
let’s stipulate that duties of rescue only arise when the harm to be a v erte d is at least as great
as the harm of ten disloc ate d fingers. And no w c onsider the f o llo wing triad of c ases:
Sma l l Differenc e 1 . T arget c an prev ent Attacker from breaking her arm without
harming him, or b y spraining fiv e of his fingers.
Sma l l Differenc e 2 . T arget c an prev ent Attacker from breaking her arm b y dis-
loc ating fi v e of his fingers or b y disloc ating ten of his fingers.
23
As clearly indic ate d b y his c omments on ibid., 1 32.
16
La rge Differenc e . T arget c an prev ent Attacker from breaking both of her arms
without h arming him, or b y spraining ten of his fingers.
Rescue tells us that the harsher def ensiv e option in Large Diff erenc e is unne c essary . But it
doesn’t sa y the same of the harsher def ensiv e options in the tw o Small Diff erenc e c ases, sinc e
the harm of fiv e spraine d fingers f alls belo w the duty-of-rescue “thresho ld ”. This means that
T a rget is permitte d to sprain ten of Attacker’s fingers if he c omes f or one of her arms on
one oc c asion and f or the other arm on another oc c asion, but that T arget is not permitte d to
sprain ten of Attacker’s fingers if he c omes f or both her arms at onc e. Attacker c an a v oid a
c ertain amount of liability simp ly b y c ombining his attacks into a single dose! This is a bizarre
imp lic ation.
The def ender of Rescue thus f ac es a dilemma: either they must reje ct c ommonsense be-
liefs about when duties of rescue arise, or else reje ct c ommonsense beliefs about the wrong-
fulness of gratuitous def ensiv e harm. Neither horn is attractiv e.
But ev en if w e are willing to grab ho ld of one of these horns, there’s a more serious prob lem
f or the ac c ount. Ac c ording to Rescue , the duty to refrain from unne c essary harm is a spe cial
instanc e of the duty to rescue. W e w ould e xpe ct, then, the duty to refrain from unne c essary
harm to ha v e the profile of a duty of rescue. But it doesn’t. Consider, f or instanc e:
Two Drivers . There are tw o re ckless driv ers, A and B. Both lose c ontro l of their
c ars. A ’s c ar c rashes into a pond, where it is quickly sinking. B’s c ar c are ens
to w ards y ou. If y ou do nothing, y ou will be kille d. Y ou c an prote ct y ourself in
one of thre e w a ys. First, y ou c an disab le B’s c ar with an ele ctromagnetic pulse,
in which c ase B will be unharme d and y ou will sa v e A from dro wning. Se c ond,
y ou c an jump out of the c ar’s p ath, in which c ase B will be unharme d but y ou
17
will be unab le to sa v e A. Third, y ou c an thro w a grenade at B’s c ar, in which
c ase B wi ll be kille d but y ou will sa v e A from dro wning.
Y ou should p lainly choose to use the ele ctromagnetic pulse. But it also p lain that thro wing
the grenade is worse than jumping out of the c ar’s p ath: y ou do a much greater wrong b y
killing B than b y f ailing to rescue A. Rescue pre dicts otherwise, ho w ev er, as it claims that
the wrong of killing A just is the wrong of f ailing to rescue A.
24
1.3 Liability, Duty, & Proportionality
The Rescue ac c ount points us in the right dire ction: it appeals to dire cte d duties to e xp lain
the ne c essity c onstraint. But it appeals to the wrong sorts of dire cte d duties. It gets things
b ackw ards. Where Rescue claims the ne c essity c onstraint is grounde d in duties of the de-
fender , I claim the c onstraint is grounde d in duties of the attac ker . The remainder of the
p aper dev elops and def ends this idea.
The ac c ount takes some time to build up to . W e w on’t be ab le to get the full story in front
of us until §1.4. W e first ne e d to unp ack the the ory of def ensiv e liability and the the ory of the
proporti ona lity c onstraint on which m y ac c ount of ne c essity is pre dic ate d. That is the proje ct
of §1.3.
24
No w it might be rep lie d that thro wing the grenade is w orse be c ause it is a f ailure to rescue some one from
harm f or which y ou are responsib le. But it is not true as a general matter that w e ha v e a spe cial ob ligation to
rescue pe op le from harm f or which w e are responsib le. Suppose A is trying to kill me, and B is trying to kill
y ou. W e def end ourselv es in the only w a y w e c an: w e each inje ct our respe ctiv e attackers with a non-lethal
poison. Untreate d, the poison will render them be dridden and in moderate disc omf ort f or a month. Y ou soon
c ome into possession of a vial of antidote, but only enough to cure one or the other of A or B. Do y ou ha v e an
ob ligation to giv e the antidote to B rather than A, in virtue of the f act that y ou bear some responsibility f or B’s
illness? It se ems not.
18
1.3.1 Defensive Liability by Duty
In §1.2 w e observ e d that unne c essary def ensiv e harm wrongs its target. This suggests a start-
ing p lac e f or a the ory of the ne c essity c onstraint: in v estigating the me chanics of rights-suspension.
If w e w ant to understand wh y an unjust attack only suspends the attacker’s rights against ne c-
essary def ensiv e harm, w e should start b y asking wh y an unjust attack suspends an attacker’s
rights at all.
I believ e the most e xp lanatorily fruitful answ er to this question is one that has be en pro-
pose d b y V ictor T adros.
25
Ac c ording to T adros, the usual w a y in which an unjust attack
brings about the loss of rights is this: the attack puts the attacker under c ertain duties, and it
is the presenc e of those duties that in turn e xp lains the attacker’s loss of rights. This claim is a
spe cific app lic ation of a more general princip le as to ho w a person’s rights ma y be moderate d
b y that person’s duties: For some duti es, t he f act t hat S has a duty to see to it t hat p ma kes it t he
case t hat S loses c erta i n rig hts ag a i nst bei ng ha rmed i n servi c e of p .
26
App lie d to def ensiv e harm spe cific ally , the idea is this. When an attacker poses an unjust
threat, he incurs c ertain duties — to include the duty to avert his t hreat . He thereb y loses
c ertain rights against being harme d in servic e of that go al.
Defensive Liability b y Duty (DLD) . When an attacker has a duty to a v ert
his threat (or w ould ha v e such a duty w ere he ab le to a v ert his threat), the at-
tacker loses c ertain rights against being harme d as a means or side-eff e ct of the
25
V ictor T adros, The Ends of Ha rm: The Mor a l Fou ndati ons of Cri mi na l Law (Oxf ord: Oxf ord Univ ersity
Press, 2011), 129; T adros, “Duty and Liability ,” Uti litas 24 (2012): 259-277.
26
The ‘some’ is signific ant here, as there are man y duties that do not make their bearers liab le to be harme d
as a means of se eing to that duty’s end. M an y duties relate d to spe cial relationships are like this. F or e x amp le,
man y of the positiv e duties that arise betw e en friends or spouses are p lausib ly non-enf orc eab le, such as duties
of gratitude and lo y alty .
19
a v ersion of his threat.
27
DLD nic ely e xp lains a number of puzzling p henomena regarding def ensiv e harm.
First, it e xp lains wh y unjust attackers “regain” man y of their rights against harm onc e their
attack c oncludes. Re ckless loses c ontro l of his c ar. It hurtles to w ards y ou. If y ou c an only
def end y ourself b y breaking Re ckless ’s leg, y ou are surely permitte d to do so . While his c ar
hurtles to w ards y ou, he lacks a right against this use of def ensiv e harm. But suppose y ou aren’t
ab le to def end y ourself. The c ar hits y ou, breaking y our leg. Re ckless stumb les out of the c ar,
daze d and c onfuse d. Y ou no w ha v e an opportunity to break his leg. M a y y ou do so? Plainly
not. Although y ou ma y ha v e bro ken Re ckless ’s leg to def end y ourself while his c ar hurtle d
to w ards y ou, y ou ma y no longer do so . DLD e xp lains w h y . What ultimately matters isn’t the
time at which y ou break Re ckless ’s leg. What matters is whet her imposing harm on Re ckless
serv es an end he has a duty to se e to . (Consider: ev en while his c ar is hurtling to w ards y ou,
y ou ma y not break Re ckless ’s leg if doing so w ould in no w a y a v ert his threat.) While his c ar
hurtles to w ards y ou, he has a duty to a v ert his threat; y ou ma y thus use def ensiv e harm that
serv es to a v ert that threat. After his c ar has hit y ou, breaking Re ckless ’s leg does nothing to
a v ert the threat he previously pose d.
Se c ond, and relate dly , DLD e xp lains wh y unjust attackers do not mak e themsel v es liab le
to be harme d f or just an y v aluab le purpose; it e xp lains wh y liability is “go al-relativ e”.
28
Con-
sider:
27
F or the sake of simp licity , I ref er to ‘attackers ’ in this f ormulation. But strictly speaking, attackers are not
the only p e op le who c an ac quire duties that make them liab le to def ensiv e harm. F or e x amp le, a person who
willingly acts as a human shield f or an unjust attacker ma y be liab le to def ensiv e harm. So too a person who
c ould easily mo v e aside to let y ou esc ape an attacker. F or e x amp les like these, se e Fro w e, Defensive Ki l li ng , 25,
and M cM ahan, Ki l li ng i n W a r , 172.
28
Se e M cM ahan, Ki l li ng i n W a r , 8-9.
20
Inci denta l Defense . Driv er risks driving in a sno wstorm and loses c ontro l: the
c ar hurtles to w ards T arget. T arget c an re dire ct the c ar to the left or right. If
she re dire cts right, she will suff er no harm but Driv er will be p aralyze d. If she
re dire cts left, her and Driv er will both be p aralyze d, but the c ar will stop an
unrelate d a v alanche from p aralyzing Bystander.
Intuitiv ely , T arget ma y not re dire ct Driv er to the left. But this is puzzling. Wh y ma y T arget
not p aralyze Driv er in order to rescue Bystander when T arget w ould be permitte d to p aralyze
Driv er in order to sa v e herse lf ? DLD giv es us an answ er. Driv er poses an unjust threat to
T arget, but not to Bystander. In virtue of this f act, Driv er has a duty to a v ert the threat of
p aralysis f acing T arget, but no duty to a v ert the threat of p aralysis f acing Bystander. Thus,
Driv er is liab le to be harme d as a means of a v erting the threat to T arget, but not liab le to be
harme d as a means of a v erting the threat to Bystander.
1.3.2 The Proportionality Constraint
A third e xp lanatory upshot of DLD is that it off ers us a w a y of making sense of the pro-
portionality c onstraint. The proportionality c onstraint re quires (roughly) that the def ensiv e
harm impose d on an unjust attacker not be “too great” relativ e to the sev erity of the threat
a v erte d. The c onstraint takes the f orm of an upper limit on the sev erity of def ensiv e harm
such that (all else e qual) the greater the sev erity of the threat a v erte d, the greater this upper
limit.
In the remainder of §1.3 I’ll argue t hat DLD e xp lains both these f eatures: it e xp lains (i) wh y
there is such an upper limit in the first p lac e, and (ii) wh y that upper limit slopes up w ards in
relation to the sev erity of the threat a v erte d. I’ll also argue that DLD e xp lains a third f eature
21
of our pre-the oretic c onc eption of the proportionality c onstraint: it e xp lains (iii) wh y that
upper limit often e x c e e ds the sev erity of the threat a v erte d — i. e., wh y it is often proportionate
to impose def ensiv e harm that is more sev ere than the threat a v erte d. Let’s unp ack these thre e
e xp lanatory upshots in turn.
W h y ther e is an upper limit . If w e refle ct on the structure of duties w e se e that they do
not just diff er with respe ct to their bearers (i. e., who has the d uty) and their ends (i. e., the
actions or go als they assign the bearer). They also diff er with respe ct to ho w demanding they
are. Suppose m y child f alls into a pond in front of a stranger; m y child will dro wn if the
stranger does not rescue her. On a sep arate oc c asion m y child again f alls into the pond, this
time in front of me; she will dro wn if I do not rescue her. On these tw o oc c asions the stranger
and I ac quire duties with similar ends — namely , to rescue m y child from the pond. But our
duties diff er in ho w demanding they are. The stranger’s duty ma y not re quire him to bear
massiv e c osts to rescue m y child. If he c an rescue m y child only b y sac rificing, sa y , both his
legs, it ma y be permissib le f or him to abstain. Not so f or me. Giv en m y spe cial relationship
to m y child, m y duties to w ards her are a great deal more demanding than a stranger’s.
The demandingness of a duty is the upper limit of c ost the agent must bear in order to
discharge that duty: if the agent c an discharge the duty only b y bearing greater c ost than this,
then it is permissib le f or her to not discharge the duty , and the person to whom that duty is
o w e d c annot reasonab ly demand that the agent discharge the duty .
29
This is an important
f eature of a duty , distinct from its bearer and its end. When w e desc ribe a duty w e might use
the f o llo wing schema to c apture all thre e f eatures: “ S has a duty , of demandi ngness d , to se e
29
Though of c ourse they ma y still demand c ompensation. W e sometimes o w e c ompensation ev en f or per-
missi b le rights infringements, as illustrate d b y Joel F einberg’s w ell-w orn c ase of the hiker who must break into
the unoc cupie d mountain c abin to sa v e his o wn lif e. Se e F einberg, “V o luntary Euthanasia and the Inalienab le
Right to Lif e,” Phi losop hy a nd Pu b li c Aff a i rs 7 (1978): 93-123.
22
to it that e .”
DLD tells us that an attacker who has a duty to a v ert his threat is thereb y liab le to def ensiv e
harm; he is liab le to be harme d as a means or side-eff e ct of the a v ersion of his threat. These
duties are like an y other duty in ha ving a c ertain lev el of demandingness: there is an upper
limit, d , to the c ost the attacker must bear to a v ert his threat. If t he c ost of a v erting the threat
is greater than d , then it is permissib le f or the attacker to not a v ert his threat, and his target
c annot demand that the attacker bear c osts greater than d to a v ert his threat.
But if the def ender c annot demand that the attacker incur c osts greater than d , then surely
the def ender c annot make the attacker bear c osts greater than d . Giv en DLD , then, it f o llo ws
that the demandingness of the attacker’s duty imposes an upper limit on ho w much c ost he
c an be made to bear in servic e of that duty .
30
Notic e that what w e’v e arriv e d at is a c onstraint that has the upper-limit structure of the
proportionality c onstraint. M y claim is that this is the proportionality c onstraint.
T he Demands of Pr opor tion ality . Def ensiv e harm is narro wly dispro-
portionate to the threat a v erte d just in c ase the sev erity of the def ensiv e harm
30
It ne e d not be the c ase that there is only one duty that grounds some one’s def ensiv e liability . It ma y be that
an attacker has multip le enf orc eab le duties that c an be enf orc e d b y imposing def ensiv e harm on him. In such
c ases, I think the upper limit of def ensiv e harm that ma y be impose d on the attacker will be e quiv alent to an
aggreg ate of the demandingness of those duties. But def ending this claim in the present p aper w ould take us
too f ar afield.
23
e x c e e ds the d emandingness of the attacker’s duty to a v ert that threat.
31
W h y this upper limit sl opes up w ar ds with se v erity of thr eat . In addition to e xp laining
wh y there is an upper limit on def ensiv e harm, DLD also e xp lains wh y that upper limit slopes
up w ards in relation to the sev erity of the attacker’s threat (all else e qual). T o se e this, let’s
c onsider a c ase where some one is indisputab ly under a duty to a v ert a threat. W e c an then
refle ct on what is independently p lausib le about the demandingness of that person’s duty .
Bott le Roc ket . Prankster, as an ill-c onc eiv e d jo ke, p lac es a bottle rocket ne xt to
V ictim while she sle eps. The bottle rocket is on a timer, and will detonate in
fiv e minutes if Prankster does not disab le it. If the rocket detonates, V ictim will
suff er harm of s ev erity n .
It is ob vious that Prankster has a duty to disab le the bottle rocket. Ho w demanding is this
duty? It surely depends on the v alue of n . Comp are tw o v ersions of the c ase. In the first,
Prankster kno ws that V ictim will ha v e both her legs sev ere d if the rocket detonates. In the
se c ond v ersion, Prankster kno ws that V ictim will only suff er a minor burn to her big toe. In
31
Comments from T adros suggest he endorses something along these lines: se e his Ends of Ha rm , 347, as
w ell as his W rongs a nd Cri mes (Oxf ord: Oxf ord Univ ersity Press, 2016), 172. It’s not clear whether he w ould
endorse the v arious claims to f o llo w in this se ction, ho w ev er. Quong also endorses a princip le in the neighbor-
hood, though he f ocuses on the demandingness of a slightly diff erent duty . On his ac c ount, proportionality
is determine d b y the demandingness of the dire cte d duty not to ha rm that w ould be vio late d b y the attacker
w ere his attack suc c essful. One reason w e might think m y princip le pref erab le to Quong’s c onc erns c ases of
“p artial def ense”. Suppose Attacker is attempting to harm T arget such that Attacker’s duty not to harm has
a demandingess lev el of 100 units of harm. Def ender is only ab l e to prote ct T arget from a tin y portion of the
attack: a mere 1%. And Def ender c an do so only b y imposing 100 units of harm on Attacker. If this much
def ensiv e harm is the v ery limit of what w ould be proportionate to a v ert the enti re threat, then surely it is dis-
proporti onate to use this much harm to a v ert only 1% of the threat. But Quong’s princip le pre dicts that this
use of def ensiv e f orc e is proportionate, sinc e it does not e x c e e d the demandingness of the duty that w ould be
vio late d b y the attacker w ere his attack suc c essful. M y Demands of Proportionality princip le does not
ha v e this result, sinc e it fix es proportionality to the demandingness of the duty to a v ert the portion of the threat
a v erte d. (Cre dit to Kida Lin f or this point.)
24
the high-stakes v ersion, it is clear that Prankster’s duty w ould be v ery demanding. E v en if
disab ling the rocket meant, f or e x amp le, that Prankster w ould lose both his legs, he w ould
still be re quire d to disab le the rocket.
Not s o in the lo w-stakes v ersion of the c ase. It is not p lausib le that Prankster w ould be
re quire d to disab le the rocket if it meant losing both his legs. M orality w ould c ertainly re quire
that Prankster c ompensate V ictim f or her minor burn, but it w ould not re quire him to giv e
up both his legs to prev ent the burn.
What these tw o c ases illustrate is that — whatev er w e should think about the relationship
betw e en duty and def ensiv e liability — the demandingness of an unjust attacker’s duty to
a v ert his threat is p artly a function of the sev erity of his threat. All else e qual, the greater the
threat, the more demanding his duty to a v ert that threat. If w e c ombine this independently
p lausib le thought with Demands of Proportionality it f o llo ws that the proportionality
limit “slopes up w ards ”: all else e qual, the greater the threat, the greater the range of def ensiv e
harm that c ounts as proportionate.
W h y pr oportionate harm c an be mor e se v er e than the thr eat a v erte d . As w e note d, it
isn’t p lausib le that Prankster w ould be morally re quire d to lose tw o legs in order to prev ent
V ictim from suff ering one mild burn; his duty w ouldn’t be this demanding. But suppose
Prankster c ould disab le the rocket at the c ost of, sa y , two mild burns to himself. It is intu-
itiv ely p lain that Prankster w ould be re quire d to disab le the bomb at such a c ost, ev en though
this c ost is greater than the c ost V ictim w ould suff er if the rocket w ere to detonate (tw o burns
v ersus one). The upshot: whatev er one thinks about the relationship betw e en duty and de-
f ensiv e liability , it is p lausib le that the demandingness of our duties to a v ert our unjust threats
is often greater than the sev erity of those threats. If w e c ombine this independently p lausib le
25
thought with Demands of Proportionality it f o llo ws that def ensiv e harm will often be
proportionate ev en when it is more sev ere than the threat a v erte d.
32
1.4 The Demands of Necessity
1.4.1 The Positive Account
Unjust attackers are liab le to def ensiv e harm be c ause they ha v e a duty to a v ert their threat (or
w ould ha v e such a duty w ere they ab le to a v ert their threat). And they are liab le only to as
mu c h harm as they w ould be re quire d to take upon themselv es to a v ert their threat. This, I
argue d in the previous se ction, is the b asic idea e xpresse d b y the proportionality c onstraint.
Def ensiv e harm is disproportionate to the threat a v erte d when the attacker w ould not be
re quire d to bear that amount of harm to a v ert that threat.
On this ac c ount, the proportionality of a def ensiv e option in v o lv es a c omp arison betw e en
tw o v ariab les. First, there is the v ariab le of “ def ensiv e gain”. This is a measure of the harm
a v erte d to persons ot her t ha n t he attac ker — the diff erenc e in ho w much harm they suff er on
the def ensiv e option as c omp are d to the harm they suff er if no def ense is mounte d (the “non-
def ensiv e option”). Se c ond, there is the v ariab le of “ def ensiv e c ost”. This is the diff erenc e in
32
I anticip ate the w orry , ho w ev er, that Demands of Proportionality “undersells ” the upper limit of
def ensiv e harm to which c ertain sorts of attackers are liab le. In p articular, mi ni ma l ly-responsi b le attackers. Con-
sider: an unf orese eab le me chanic al f ailure c auses a c areful driv er to lose c ontro l of his c ar, threatening to kill a
pe destrian (se e M cM ahan, Ki l li ng i n W a r , 165). It is widely believ e d that a def ender w ould act permissib ly
and proportionately b y killing the driv er in def ense of the pe destrian. But — some might w onder — w ould the
driv er really be re quire d to re dire ct his c ar a w a y from the pe destrian, and to his o wn death, w ere he ab le to do
so? Y es, I think he w ould be. When w e’re dealing with an attacker’s duty of threat a v ersion, w e’re not dealing
with a generic duty of rescue. The duty to a v ert one’s attack is an e xtension of one’s more fundamental neg ative
duty not to harm. And negativ e duties are quite demanding — so demanding that a duty not to kill c an ev en
demand one’s lif e. If a villain promises to kill me unless I kill an innoc ent stranger, this doesn’t make it permis-
sib le f or me to kill the stranger. It doesn’t matter that m y r esponsibility f or the killing w ould be signific antly
mitigate d b y m y o wn duress. I am re quire d to discharge m y duty not to kill ev en at the c ost of m y lif e. So too is
the c onscientious driv er re quire d to discharge his duty not to kill ev en at the c ost of his lif e.
26
ho w much harm t he attac ker suff ers on the def ensiv e option as c omp are d to the non-def ensiv e
option. T o illustrate: Suppose Attacker will kill fiv e pe op le if Def ender doesn’t act; but if
Def ender shoots and kills Attacker, no one else will be harme d. In this c ase, the def ensiv e
gain of shooting Attacker is the sa v e d liv es of his fiv e targets. The def ensiv e c ost is the loss of
Attacker’s lif e.
The proportionality c onstraint boils do wn to the f o llo wing test of a def ensiv e option:
W ou ld t he attac ker be requi red to suffer t he defensive c ost of t hat opti on to secu re its defensive
g a i n? If Y es, the option is proportionate. If No , the option is disproportionate.
The proportionality c onstraint thus in v o lv es a c omp arison betw e en a giv en def ensiv e op-
tion and the non-def ensiv e option. But there’s nothing morally privilege d about the non-
def ensiv e option. A def ender ma y ha v e a number of options a v ailab le to her, and if this sort
of c omp arativ e test matters with respe ct to the non-def ensiv e option, then it should also mat-
ter with respe ct to the def ender’s ot her options. Suppose the def ender is c onsidering def ensiv e
option D
1
, but has another option a v ailab le to her, D
2
. If the attacker w ould not be mo rally
re quire d to bear the e xtra c ost impose d on him b y D
1
as c omp are d to D
2
in order to se cure
the e xtra def ensiv e gains of D
1
as c omp are d to D
2
, then the attacker is not liab le to the c osts of
D
1
. It makes no diff erenc e whether D
2
is the non-def ensiv e option or some other def ensiv e
option.
This, I claim, is the idea at the heart of the nec essity c onstraint. The ne c essity c onstraint is
just like the proportionality c onstraint in the sort of c omp arativ e test it in v o lv es. But where
the proportionality c onstraint tests a def ensiv e option against a single alternativ e (the non-
def ensiv e option), the ne c essity c onstraint tests a def ensiv e option against every alternativ e
option a v ailab le to the def ender.
27
T he Demands of N ecessity . A def ensiv e option is unne c essary just in c ase
there is an alternativ e def ensiv e option such that the attacker w ould not be re-
quire d to bear the diff erenc e in def ensiv e harm betw e en the tw o options to se-
cure the d iff erenc e in def ensiv e gain.
This princip le (‘ Necessity ’ f or short) deliv ers a s ymmetric e xp lanation of both “pure”
c ases of unne c essary def ense (like E as y Def ense) and “marginal ” c ases (like E as yish Def ense).
Consider, first, the c ase of E as y Def ense: P e destrian c an e qually w ell def end herself b y p ara-
lyzing Re ckless or b y spraining his toe. It is p lainly wrongful to p aralyze him. Here’s wh y ,
ac c ording to Necessity . Necessity tells us to make a p airwise c omp arison betw e en the
option to p aralyze Re ckless and ev ery other a v ailab le option. In this c ase, the option to p ara-
lyze Re ckless f ails the relev ant test with respe ct to the option to sprain Re ckless ’s toe: Re ckless
w ould p lainly not be re quire d to bear the diff erenc e be tw e en p aralysis and a spraine d toe in
order to bring about zero def ensiv e gain. This is wh y P e destrian w ould wrong Re ckless b y
p aralyzing him.
Likewise, c onsider E as yish Def ense: P e destrian will a v oid harm entirely if she p aralyzes
Re ckless, but she will suff er a spraine d toe if she def ends herself b y spraining Re ckless ’s toe.
Necessity dire cts us to ask whether Re ckless w ould be re quire d to bear the diff erenc e be-
tw e en p aralysis and a spraine d toe — a massiv e diff erenc e — in order to prote ct P e destrian
from the minor harm of a spraine d toe. Plainly he w ould not be, as w e c an se e if w e imagine a
c ase in which the i nitia l threat from Re ckless ’s c ar is nothing more than a spraine d toe. The
v ery reason it w ould be disproporti onate f or P e destrian to p aralyze Re ckless when the initial
threat is so small is the same reason it w ould be u n nec essa ry f or P e destrian to p aralyze Re ck-
less rather than sprain his toe in E as yish Def ense: Re ckless w ould not be re quire d to bear the
28
massiv e c ost of “almost-p aralysis ” in order to a v ert the small harm of a spraine d toe.
33
1.4.2 Advantages of the Account
This w a y of interpreting the ne c essity c onstraint has signific ant upsides. First, Necessity
off ers an impro v ement o v er the Rescue ac c ount. Whereas Rescue (I argue d) pro hibits
only large amounts of gratuitous def ensiv e harm, Necessity pro hibits a n y amount. This
is be c ause the duties that typic ally render pe op le liab le to def ensiv e harm are duties to avert
some t hreat .
34
Def ensiv e harm that is gratuitous — that f ails to a v ert an y threat — is, of c ourse,
harm that f ails to enf orc e an y such duty . The upshot is not just that Necessity is a better fit
with our intuitiv e judgments about p articular instanc es of gratuitous harm, but also that it
a v oids c ounterintuitiv e f ailures of harm additivity that are generate d b y the thresho ld f eature
of Rescue .
Necessity also off ers an impro v ement o v er Discount . The biggest impro v ement is t hat
it is c onsistent with the f act that (and e xp lains wh y) unne c essary harm wrongs an attacker.
But Necessity is also pref erab le in that it assigns a more p lausib le ro le to the presenc e of
b ystanders. Re c all that Disc ount runs into troub le with the f o llo wing p air of c ases:
Un nec essa ry Ki l li ng . Re ckless loses c ontro l of his c ar, which will permanently
p aralyze P e destrian from the w aist do wn unless she either kills Re ckless or merely
33
M cM ahan, “Limits of Self-Def ense,” 189-191 claims that c alculations of ne c essity c an be transf orme d into
c alculation of wi de proportionality . By c ontrast, the Necessity ac c ount claims that c alculations of ne c essity
c an be transf orme d into a set of c alculations of na rrow proportionality (giv en the ac c ount of proportionality
def ende d in §1.3). If y ou w ant to kno w whether def ensiv e option D
1
is ne c essary , run the f o llo wing test f or ev ery
alternativ e def ensiv e option, D
n
: imagine that Attacker’s i nitia l threat is the diff erenc e betw e en his threat in D
1
and his threat in D
n
, and then ask y ourself whether it w ould be proporti onate to make him bear the diff erenc e in
def ensiv e harm betw e en D
1
and D
n
to a v ert t hat threat. If there is some option f or which this proportionality
test f ails, D
1
is unne c essary .
34
I sa y ‘typic ally’ f or reasons I will u np ack in chapter 2.
29
sprains one o f his fingers.
Un nec essa ry Ki l li ng (Bysta nders) . As bef ore, e x c ept that killing Re ckless in-
v o lv es re dire cting his c ar into the p ath of an unrelate d a v alanche. The a v alanche
will kill Re ckless, but his c ar will b lock the a v alanche from killing tw o innoc ent
b ystanders.
Killing Re ckless in the first c ase is a p aradigm vio lation of the ne c essity c onstraint. But
it is surely wrong to kill Re ckless in the se c ond c ase f or the v ery same reasons it is wrong to
kill him in the first c ase. Discount, ho w ev er, deliv ers an as ymmetric v erdict in these tw o
c ases. It tells us that killing Re ckless is unne c essary in the first c ase, but not the se c ond. This
is imp lausib le.
The Necessity ac c ount does better here. This is be c ause, in bot h c ases, Re ckless w ould
not be re quire d to let himself be kille d (rather than suff ering a spraine d finger) in order to
se e to the def ensiv e gain of his being kille d. In the first c ase, there is nothing to be gaine d b y
killing Re ckless. Re ckless w ould p lainly not be re quire d to let himself be kille d f or nothing.
In the se c ond c ase, there is something to be gaine d b y killing Re ckless: tw o b ystanders will
be sa v e d from death. But Re ckless, unrelate d as he is to the threat to the b ystanders, has no
spe cial ob ligations to them. His duties to w ards the b ystanders are the same duties of rescue
that an y one else w ould ha v e to them. And our duties of rescue are not so demanding as to
re quire us to giv e up our liv es to rescue tw o pe op le from death. Re ckless, ac c ordingly , is not
re quire d to let himself be kille d in order to se e to it that tw o b ystanders are sa v e d. This is wh y
the Nec essity ac c ount de clares it to be unne c essary to k ill Re ckless ev en in the se c ond c ase.
This isn’t to sa y that Necessity treats b ystanders as making no diff erenc e. Bystanders
c an aff e ct whether def ensiv e harm is ne c essary pre cisely be c ause attackers c an o w e things to
30
persons other than their intende d victims. W e se e this clearly when w e c onsider a c ase like:
Dodge t he Bou lder . Intending to break T arget’s leg, Attacker ro lls a boulder
to w ards her. But T arget jumps out of the w a y , making it so that the boulder
no w threatens to break Bystander’s leg. Bystander c an be sa v e d only if Attacker
thro ws hims elf in front of the boulder, at the c ost of his o wn bro ken leg.
Attacker is intuitiv ely re quire d to suff er a bro ken leg in def ense of Bystander — this despite
the f act that he ma y not ha v e intende d or f orese en harm to Bystander, and despite the f act
that Bystander is threatene d in p art be c ause of a c hoi c e made by T a rget .
In the same w a y , def ensiv e harm that w ould be unne c essary in the absenc e of b ystanders
c an be made ne c essary b y their presenc e. But this is pre cisely the ro le w e should e xpe ct b y-
standers to p la y , and is c onfirme d b y refle ction on c ases. Consider, f or e x amp le, a p air of
v ariants of Dodge the Boulder:
Defensive Bou lder 1 . Attacker ro lls a boulder to w ards T arget in order to break
her leg. T arget c an thro w Attacker in front of the boulder, breaking his leg, or
she c an dodge the boulder without harm to an y one. T arget chooses to thro w
Attacker i n front of the boulder.
Defensive Bou lder 2 . As bef ore, e x c ept that if T arget dodges the boulder, it will
break Bystan der’s leg. T arget thro ws Attacker in front of the boulder.
Def ensiv e Boulder 1 is a p aradigm c ase of impermissib le and unne c essary def ensiv e harm.
And y et the v ery same def ensiv e harm — breaking Attacker’s leg — is permissib le and ne c es-
sary in Def ensiv e Boulder 2. The presenc e of Bystander changes the status of T arget’s action
31
from unne c essary to ne c essary , from impermissib le to permissib le. Necessity e xp lains wh y .
The presenc e of Bystander makes it permissib le f or T arget to thro w Attacker in front of the
boulder be c ause the presenc e of Bystander means that there is something to be gaine d b y
doing so — a gain that Attacker w ould be ob ligate d to bear the c ost of (as w e estab lishe d in
Dodge the Boulder). Necessity , then, does allo w the presenc e of b ystanders to make a dif-
f erenc e as to whether def ensiv e harm is unne c essary . But the ro le it assigns to b ystanders is
much more p lausib le than the ro le Discount assigns them.
A v oiding the shortc omings of c ompeting ac c ounts is one virtue of Necessity . But per-
haps its greatest virtue is the unity it brings to the ethics of def ensiv e harm. Necessity helps
us to se e that the so-c alle d “ne c essity” and “proportionality” c onstraints are just tw o f ac es of
a c ommon moral demand. Both c onstraints demand of a def ensiv e option, relativ e to c ertain
alternativ e options, that the c omp arativ e c osts to the attacker be ones he w ould be re quire d
to bear in order to se e to the c omp arativ e def ensiv e gains. The tw o traditional c onstraints dif-
f er only in the c omp arators in v o lv e d. The proportionality c onstraint c omp ares a def ensiv e
option with the non-def ensiv e option; the ne c essity c onstraint c omp ares a def ensiv e option
with every alternativ e option a v ailab le to the def ender. The tw o c onstraints impose a c om-
mon test, but the ne c essity c onstraint imposes the test more bro adly . Disproportionate harm,
it turns out, is just a spe cial instanc e of unne c essary harm.
1.5 Necessity as Responsibility-Sensitive
In the remainder I c onsider what f o llo ws from Necessity . In this se ction I argue that the
ac c ount has important imp lic ations f or the relationship betw e en responsi bi lity and ne c essity .
In the f o llo wing se ction I take up the w orry that the ac c ount has a troub ling imp lic ation in
32
c ases in v o lving multip le attackers.
M an y the orists ha v e claime d that an attacker’s degre e of responsibility f or his threat aff e cts
what c ounts as proportionate def ensiv e harm — that (all else e qual) the greater the attacker’s
responsibility , the greater the degre e of def ensiv e harm it is proportionate to impose.
35
One
important imp lic ations of the p ackage of claims I’v e def ende d in §1.3 and §1.4 is that an
attacker’s degre e of responsibility (or culp ability) aff e cts bot h what c ounts as proportionate
def ensiv e harm and what c ounts as ne c essary def ensiv e harm.
36
This is be c ause it is indepen-
dently p lausib le that all else e qual, the more responsib le a person is f or posing an unjustifie d
threat, the more demanding his duty to a v ert that threat. Consider the f o llo wing p air of c ases:
Sleepy Condu ctor . Sam arriv es to her job as a train c onductor with v ery little
sle ep , ha ving sta y e d up most the night to binge her f a v orite sho w . Her lack of
sle ep c auses her to f ail to notic e a w orker who has be en b lo wn onto the tracks.
Sam c an only stop the train from breaking the w orker’s ankles b y thro wing her-
self in fro nt of the train. This will result in n -much harm to S am.
V engefu l Condu ctor . Sully is a disgruntle d train c onductor. T o get b ack at the
rail c omp an y f or p assing him o v er f or promotion, he pushes a w orker onto the
tracks. Sully c an only stop the train from breaking the w orker’s ankles b y thro w-
ing hims elf in front of the train. This will result in n -much harm to Sully .
Sam and Sully each ha v e a duty to a v ert the threats f or which they are responsib le. But it
35
Se e, f or e x amp le, Sab a Bazargan-F orw ard, ”Killing Minimally Responsib le Threats,” Et hi cs 125 (2014):
114-136; Kaila Draper, ”Def ense,” Phi losop hi ca l Stu di es 145 (2009): 81; Rodin, “Justifying Harm,” 80-84;
T adros, Ends of Ha rm , 332; and Suzanne Uniacke, “Proportionality and Self-Def ense,” Law a nd Phi losop hy 30
(2011): 265.
36
M y o wn view is that responsibility and culp ability make independent and distinct c ontributions to the de-
mandingness of an attacker’s duty . Se e chapter thre e (§3.4.3) f or a brief discussion of this point. F or simp licity ,
though, I will ref er only to responsibility throughout this se ction.
33
se ems p lain that Sully’s duty is much more demanding than Sam’s: the range of v alues of n
such that Sully is morally re quire d to thro w himself in front of the train is much greater than
the range of v alues of n such that Sam is morally re quire d to thro w herself in front of the
train. This much is intuitiv ely clear, ev en if our intuitions are hazy as to e x actly what those
tw o ranges are.
Sully’s duty isn’t more demanding be c ause of a diff erenc e in the harmfulness of the threat;
in both c ases the threat to the w orker is the same. Nor is Sully’s duty more demanding be c ause
of a diff erenc e in the justific ation of the threat; in both c ases the threat is unjustifie d. The
relev ant diff erenc e c an only be Sam and Sully’s c omp arativ e degre es of responsibility . Sully’s
duty is more demanding be c ause he bears greater responsibility f or the threat.
The general lesson is that a person’s responsibility f or an unjust threat is a c ontributing
f actor to the demandingness of their duty to a v ert that threat. All else e qual, the greater y our
responsibility f or an unjust threat, the greater y our duty to a v ert that threat. Giv en Demands
of Proportionality , this means that (all else e qual) the greater an attacker’s responsibility
f or an unjust threat, the greater the range of def ensiv e harm that c ounts as proportionate.
This se ems right on refle ction. Imagine that Sam and Sully both refuse to thro w themselv es
in front of the train, but that a third p arty c ould push Sam and Sully in front of the train. It
is v ery intuitiv e that Sully is liab le to a greater range of harm from such a push than is Sam.
What is perhaps more surprising is that w e get similar imp lic ations regarding nec essity . If
greater responsibi lity makes f or more demanding duties, then it f o llo ws from Necessity
that (all else e qual) the greater an attacker’s responsibility f or an unjust threat, the more de-
f ensiv e options will c ount as ne c essary . W e c an se e this with the f o llo wing to y e x amp le. Sup-
pose that the demandingness of Sam the Sle ep y Conductor’s duty to a v ert harm to the w orker
34
is identic al to the sev erity of the harm she c an a v ert, whereas the demandingness of Sully the
V engeful Conductor’s duty is twi c e as great as the harm he c an a v ert.
And no w suppose Def ender has the f o llo wing options f or ho w she might push Sam and
Sully onto their respe ctiv e tracks:
Opti on 1 : Def ender does not hi ng . W orker suff ers 2 bro ken ankles.
Opti on 2 : Def ender giv es Sam/Sully a gent le pus h . W orker suff ers only 1 bro ken
ankle. Sam/Sully suff ers only 1 bro ken ankle.
Opti on 3 : Def ender giv es Sam/Sully a hefty pus h . W orker suff ers no harm. Sam/Sully
suff ers 2 bro ken ankles a nd a bro ken toe.
Giv en our stipulations, Necessity tells us that Sully is liab le to a hefty push, but that
Sam is liab le only to a gentle push. Sully w ould be re quire d to bear the diff erenc e betw e en
a hefty push and a gentle push (one e xtra bro ken ankle and one bro ken toe) in order to sa v e
the w orker from one bro ken ankle. Not so f or Sam; she w ould not be re quire d to bear an y
more than one e xtra bro ken ankle to sa v e the w orker from one bro ken ankle.
The lesson isn’t that m y the ory of proportionality and ne c essity should lead us to endorse
the so-c alle d “M oral Responsibility Ac c ount” of liability to def ensiv e harm.
37
M y ac c ount
does not pressure us to endorse the claim that an attacker is liab le to def ensiv e on ly if he is re-
37
Proponents of the claim that responsibility is ne c essary f or def ensiv e liability include Jeff M cM ahan, “The
Basis of M oral Liability to Def ensiv e Killing,” Phi losop hi ca l Issu es 15 (2005): 386-405; M cM ahan, Killing in
W ar, 155-202; Fro w e, Defensive Ki l li ng , 72-87; Kerah Gordon-So lmon, “What M akes a P erson Liab le to De-
f ensiv e Harm?,” Phi losop hy a nd Phenomeno logi ca l Resea rc h 97 (2018): 543-567; Michael Otsuka, ”Killing the
Innoc ent in Self-Def ense”, Phi losop hy a nd Pu b li c Aff a i rs 23 (1994):74-94; and Otsuka, “The M oral Responsi-
bility Ac c ount of Liability to Def ensiv e Killing”, in The Et hi cs of Se lf-Defense , e ds. Christian Coons and Michael
W eber (Oxf ord: Oxf ord Univ ersity Press, 2016): 51-68.
35
sponsib le f or the threat to be def ende d against.
38
But it does pressure us to endorse the claim
that responsibility pushes outw ards on the limits of liability . Whether or not responsibility
is ne e de d to make us liab le to def ensiv e harm in the first p lac e, it is something that c an expa nd
our liability .
1.6 Multiple Attackers
The Necessity ac c ount ma y se em to ha v e v ery troub ling imp lic ations in c ases in v o lving
multip le attackers. Consider:
Dua l Threat . Alic e and Bob each attempt to attack Def ender. If Alic e suc c e e ds,
she will break Def ender’s left leg. If Bob suc c e e ds, he will break Def ender’s right
leg. Def ender has thre e w a ys to a v ert both these threats. She c an shoot bot h Ali c e
a nd Bo b , breaking one leg each; shoot just Ali c e , breaking one of Alic e’s legs and
sc aring off Bob unharme d; or she c an shoot just Bo b , breaking one of B ob’s legs
and sc aring o ff Alic e unharme d.
It is of c ourse unne c essary to shoot bot h Alic e and Bob , sinc e shooting just one of them will
do the prote ctiv e job . But what’s w orrisome f or Necessity is that it se ems to imp ly that it is
also unne c essary to shoot just Alic e and unne c essary to shoot just Bob . This is be c ause Alic e
w ould not be re quire d to bear to c ost of a bro ken leg in order that Bo b not suff er a bro ken leg
(and vic e v ersa). The option to shoot just Alic e and the option to shoot just Bob thus se em
38
On the c ontrary , m y ac c ount pressures us to endorse the claim that responsibility f or c ontributing to a
threat is not ne c essary f or liability to def ensiv e harm. This f or the simp le reason that w e c an ac quire duties to
a v ert threats f or which w e are not responsib le. A person who positions themselv es ne ar an unjust attacker to
deter a def ender from shooting b ack ma y make themselv es liab le to def ensiv e harm ev en if they do not c ontribute
to the attacker’s threat.
36
to f ail Necessity ’s c omp arativ e test with respe ct to one another. But this w ould mean that
every def ensiv e option a v ailab le to Def ender is unne c essary , and thus that he c annot def end
himself without wronging one (or both) of his attackers. This is wildly imp lausib le.
39
The good news f or Necessity is tha t it does not, in f act, ha v e this imp lic ation. T o se e wh y ,
let’s start b y noticing that our tw o attackers, Alic e and Bob , ha v e at least some ob ligations to
one another. W e c an se e this when w e c onsider a v ariant of the c ase:
Divisi b le Dua l Threat . As bef ore, e x c ept Def ender has y et another def ensiv e
option: he c an shoot rubber bullets at Alic e and Bob , in which c ase they will
each suff er a harm e x actly half as sev ere as a bro ken leg.
Intuitiv ely , Def ender should take this option. Rather than break one attacker’s leg, he
should distribute the def ensiv e harm betw e en both attackers. What’s more, it se ems that
Alic e and Bob w ould be re quire d to sp lit the c osts in this w a y if they had the ability to do so
themselv es. Suppose Alic e c ould a v ert the threats to Def ender b y either re dire cting the full
c ost of a bro ken leg on Bob , or b y se eing to it that she and Bob each suff er half this c ost. Alic e,
I think, w ould be re quire d to choose the latter option. ( Mutatis muta ndis f or Bob .)
But notic e what this means. When w e c omp are the sp lit-t he-c osts option to the s hoot Bo b
option, it turns out that Alic e w ould be re quire d to bear half a bro ken leg in order to sa v e
Bob from half a bro ken leg. ( Mutatis muta ndis f or Bob .)
39
This w orry is diff erent from a more f amiliar prob lem in v o lving multip le attackers. The more f amiliar w orry
takes the f orm of a puzzle: Wh y does there se em to be an upper limit on the number of attackers it is permissib le
to kill in self-def ense, ev en when each of those attackers se ems to satisfy the c onditions of liability to def ensiv e
harm? F or discussion, se e Bazargan-F orw ard, “Minimally Responsib le Threats,” 133-135; Jeff M cM ahan, “Li-
ability , Proportionality , and the Number of Aggressors ”, in The Et hi cs of W a r , e ds. Sab a Bazargan and Samuel
Rickless (New Y ork: Oxf ord Univ ersity Press, 2017): 3-27; and Rodin, “Justifying Harm,” 99-100.
37
Ha ving estab lishe d that Alic e and Bob w ould o w e each other at least this much, return to
the original Dual Threat c ase. In the original c ase, the def ensiv e harm of a bro ken leg is not
divisib le. But if Alic e and Bob are ob ligate d to divide the c osts in the divisib le c ase, then it
se ems they are ob ligate d to do at least the f o llo wing in the original c ase: they are ob ligate d
to ac c ept t he c osts of a f a i r proc edu re on Defender’s pa rt for c hoosi ng whet her to s hoot Ali c e or
Bo b . In the non-divisib le c ase, this is as close as Alic e and Bob c an c ome to sp litting the c ost.
Suppose Def ender flippe d a c oin to choose betw e en shooting Alic e or Bob , which dire cte d
her to shoot Alic e. Giv en what w e’v e se en, it se ems that Alic e w ould then be re quire d to
ac c ept this c ost.
The upshot, then, is this. Necessity allo ws Def ender to prote ct herself in Dual Threat,
so long as her de cision f or choosing betw e en shooting Alic e and shooting Bob is a f air one.
So long as her de cision is not unf air, her target really does ha v e a duty to ac c ept the c osts that
f all on them. But if her de cision is unf air, no such duty is present. If, f or e x amp le, Def ender
chose to shoot Alic e rather than Bob simp ly be c ause she pref ers to shoot w omen o v er men,
Def ender w ould wrong Alic e. But this is a v ery p lausib le result. Def ender doesn’t o w e it
to Alic e or Bob not to harm them, but she does o w e it to them to giv e their interests e qual
c onsideration in her de cision of who to harm.
1.7 Conclusion
The shape of an attacker’s liability to suff er def ensiv e harm is determine d b y the s hape of his
duties: what harm an attacker is liab le to is a function of both the ends of his duties and of
ho w dema ndi ng those duties are. The ends of his duties determine that f or which def ensiv e
harm must be a means or side-eff e ct. And the demandingness of the relev ant duties tell us
38
how mu c h harm he is liab le to suff er: an attacker is only liab le to be made to bear as much c ost
in servic e of some end as he w ould be re quire d to take upon himself in servic e of that end.
I’v e claime d that w e c an understand the wrongfulness of both disproportionate and un-
ne c essary harm in terms of this f act. T o sa y that a def ensiv e option is d isproportionate is to
sa y that, when w e c omp are that option against the non-def ensiv e option, the attacker w ould
not be re quire d to assume the c omp arativ e c osts to se e to the c omp arativ e gains. T o sa y that
a def ensiv e option is unne c essary is to sa y that, when w e c omp are that option against ev ery
alternativ e option, there is some alternativ e such that the attacker w ould not be re quire d to
assume the c omp arativ e c osts to se e to the c omp arativ e gains. Thus, while there is a diff erenc e
in sc ope betw e en the proportionality and ne c essity c onstraints, there is no diff erenc e in the
fundamental nature of the wrongfulness of disproportionate and unne c essary harm. Both
f orms of harm wrong an attacker in the same w a y: b y making him to bear more c ost in servic e
of some def ensiv e go al than he w ould be re quire d to take upon himself. The tw o c onstraints
e xpress this same fundamental princip le, only at diff erent lev els of generality .
40
40
Be c ause the diff erenc e betw e en the tw o c onstraints is only a diff erenc e in sc ope, it turns out that, though it
ma y be practic ally useful to maintain the tradition of distinguishing betw e en ‘proportionality’ and ‘ne c essity’,
w e don’t really ne e d this distinction at the the oretic al lev el. What Demands of Proportionality e xpresses
in p art, Demands o f Necessity e xpresses in full generality . W e c ould, then, make do with just the latter in
our the ory of the ethics of def ensiv e harm.
39
2
Chapter T w o: Refusing Prote ction
2.1 Introduction
Refusal sometimes has the po w er to render prote ction impermissib le. Consider, f or e x amp le:
Si b li ngs : Adam, T eresa, and Dan are adult sib lings. Jealous about being left out
of their mother’s will, Adam attempts to break T eresa ’s leg. The attack c an only
be stoppe d only if either T eresa or Dan break Adam’s leg first. Dan prep ares to
40
harm his brother, but T eresa — who is a c ommitte d p acifist and who w ould
rather be harme d than se e Dan harm their brother in her name — demands
that he sta nd do wn.
Absent T eresa ’s refusal of prote ction, I think Dan w ould be permitte d to break Adam’s
leg. But giv en that she has refuse d prote ction, it se ems that Dan is not permitte d to interv ene.
He must stand do wn.
1
In at least some c ases, then, refusal se ems to render otherwise-permissib le prote ction im-
permissib le. The f o llo wing essa y is an attempt to vindic ate this intuition b y dev eloping and
def ending an e xp lanation as to wh y refusal has this po w er. I argue that the po w er to f orbid
prote ction is an upshot of a more b asic normativ e po w er — the po w er to mitigate spe cific
normativ e eff e cts of our rights. W e ha v e man y po w ers o v er our rights. W e ha v e, f or instanc e,
po w ers to transf er or w aiv e our rights. But in addition to ha ving these more f amiliar po w ers,
I claim, w e ha v e the po w er to sele ctiv ely restore some of the rights that one’s attac ker f orf eits
b y threatening our rights. One w a y in which this po w er is e x ercise d is b y the refusal of pro-
te ction. When T eresa ’s refuses Dan’s prote ction, f or e x amp le, T eresa restores Adam’s rights
that Da n not harm him — a right that Adam had lost as a result of his threat to T eresa ’s rights.
The restoration of this right is what undermines Dan’s permission to prote ct T eresa. I c all
1
M an y p hilosop hers share this intuition. F or discussion, se e: Josep h Bo w en, “Humanitarian Interv en-
tion, Other-Def ense, and Consent,” Oxford Stu di es i n Po liti ca l Phi losop hy , V o l. 8 (Oxf ord Univ ersity Press,
f orthc oming); Cé cile F abre, “P ermissib le Rescue Killings,” Proc eedi ngs of t he A ristote lia n Soci ety 109 (2009):
159-160; Christop her Finla y , “Legitimac y and Non-State P o litic al V io lenc e,” Jou rna l of Po liti ca l Phi losop hy
18 (2010): 290-295; Seth Lazar, “ Authorization and the M orality of W ar,” Austr a lasia n Jou rna l of Phi losop hy
94 (2016): 211-226; Jeff M cM ahan, The Et hi cs of Ki l li ng: Pro b lems at t he Ma rgi ns of Life (Oxf ord Univ er-
sity Press, 2002), 415; Jeff M cM ahan, “Humanitarian Interv ention, Consent, and Proportionality ,” in Et hi cs
a n d Hu ma nity: Themes from t he Phi losop hy of Jonat ha n Glover , e ds. N. Ann Da vis, Richard Keshen & Jeff
M cM ahan (Oxf ord, 2009), 49; Jonathan P arry , “Def ensiv e Harm, Consent, and Interv ention,” Phi losop hy a nd
Pu b li c Aff a i rs 45 (2017): 356-396; V ictor T adros, The Ends of Ha rm (Oxf ord 2011), 2 95-296; F ernando T esón,
Hu ma nita ria n Interventi on , 3rd e d. (Ardsley , NY : T ransnational, 2005), 160.
41
this the Liability Contraction Ac c ount.
The p aper proc e e ds as f o llo ws. §2.2 sketches tw o general strategies f or e xp laining ho w re-
fusal undermines the permissibility of prote ctiv e interv ention: either refusal undercuts the
permissibility of prote ction at the “reasons dimension” or it does so at the “liability dimen-
sion”. §2.3 c onsiders the most dev elope d attempt to e xp lain the magic of refusal at the rea-
sons dimension — Jonathan P arry’s Exclusion Ac c ount. I argue that w e should reje ct this
ac c ount, and in §2.4 I dev elop an ac c ount — the Liability Contraction Ac c ount —
ac c ording to which refusal w orks its magic at the liability dimension. §2.5 unp acks some im-
portant f eatures and imp lic ations of the ac c ount. §2.6 c onsiders the spe cial c ase of refusing
prote ction from agents of the state.
As w e’ll se e, ho w w e understand the po w er of refusal has important imp lic ations f or the
ethics of humanitarian interv ention. In real-w orld c ases, a population that is threatene d b y
an unjust aggressor will often f ail to be of one mind about prote ctiv e military interv ention
from a third-p arty nation. Some members of the population will w elc ome prote ction; some
will reje ct it. Does it make a diff erenc e what portion or number of the population reje cts
prote ction? Some the orists ha v e argue d that an ything less than the unanimous refusal of
prote ction is irrelev ant.
2
At the other end of the spe ctrum, some ha v e argue d that refusal b y
ev en a minority of the target population c an undermine the permissibility of interv ention.
The Liability Contraction Ac c ount takes a position betw e en these tw o e xtremes. It
does not treat less-than-unanimous refusal as strictly irrelev ant to the ethics of w ar. But, as
c omp are d with an ac c ount like P arry’s, Liability Contraction pre dicts a much smaller
2
Se e Altman and W ellman, “ Assassination,” 242-245; F abre, “Cosmopo litanism,” 972-974; F abre, “Rescue
Killings,” 159n11; Fro w e, “Judging Interv ention,” 108-109; and M cM ahan, “Humanitarian Interv ention,” 51-
53.
42
range of c ases in which refusal b y a subset of the population makes the diff erenc e (more on
this in §2.5).
2.2 Two Ingredients for Permissible Defen se
By refusing prote ction, the target of aggression c an sometimes render otherwise-permissib le
prote ction impermissib le. The question is ho w e x actly refusal manages to do this.
Here is a useful p lac e to begin answ ering this question. W e start b y getting bef ore us the
ingre dients that are normally re quire d f or permissib le prote ctiv e interv ention. W e then c on-
sider whether refusal undercuts this permission b y remo ving one of these ingre dients.
In p aradigm c ases of permissib le interv ention, there are tw o c entral ingre dients. The first
is attacker lia bi lity . Absent a “lesser-evil ” justific ation, a person ma y impose def ensiv e harm
on an attacker only if the attacker is liab le to suff er such harm — that is, only if the attacker
lacks a right against such harm.
3
Liability is not sufficient f or permissib le interv ention, ho w ev er. T o sa y that some one lacks
a right against some harm is only to sa y that a c ertain de ontic obstacle has be en remo v e d - that
a c ertain reason not to harm him is absent. This is not y et to sa y that there is good enough
reason to harm him. This, then, is the se c ond ingre dient: there must be suffi ci ent positive
reason to harm the attacker.
These, then, are the tw o standard c onditions on permissib le def ensiv e harm: the liability
3
M ost the orists w orking on the ethics of def ense endorse the possibility of lesser-evil justific ations f or harm-
ing non-liab le persons. These are c ases where the reasons f or imposing harm are so w eighty as to justify the in-
fringement of some one’s rights against harm. A classic e x amp le: y ou re dire ct the tro lley a w a y from killing fiv e
pe op le and onto a p ath where it will break one person’s f oot. A smaller number of the orists also endorse the
possibility of “agent-relativ e-prerogativ e” justific ations f or harming non-liab le persons — most prominently ,
Jonathan Quong, The Mor a lity of Defensive Forc e (Oxf ord, 2020), 58-96. F or simp licity , I set aside the possi-
bility of agent-relativ e justific ations in what f o llo ws.
43
c ondition and the reasons c ondition. Thus there emerge tw o strategies f or e xp laining the
po w er of refusal. One possibility is that refusal w orks its magic at the liability dimension:
it renders prote ction impermissib le b y remo ving or re ducing the attacker’s liability to suff er
def ensiv e harm. Another possibility is that refusal w orks its magic at the reasons dimension:
it renders prote ction impermissib le b y making it so that the def ender lacks “good enough”
reason to harm the attacker (either be c ause it takes a w a y reasons to def end, or be c ause it c on-
tributes c ounterv ailing reasons not to def end).
On first glanc e, the appeal to liability se ems a non-starter. Ac c ording to popular the ories of
def ensiv e liability , an attacker is liab le to suff er def ensiv e harm H when the attacker threatens
(or is responsib le f or threatening) to infringe some one’s rights and the imposition of H on
the attacker is proportionate to , and ne c essary to a v ert, that threat.
4
Giv en such a the ory , it
ma y se em that the refusal of prote ction will suc c e e d in undermining an attacker’s liability
only if it makes it so that the attacker no longer threatens some one’s rights. But there is only
one w a y in which the refusal of prote ction c an reliab ly make it so that the attacker no longer
threatens his target’s rights — which is b y making it so that those threatene d rights disappear.
The idea is that b y refusing prote ction from an attacker, a person e x ercises her po w er to w aiv e
her rights. By refusing prote ction, a target w aiv es those of her rights that are threatene d b y her
attacker, thereb y rendering her attacker non-liab le to def ensiv e harm, and thereb y rendering
prote ctiv e interv ention impermissib le. Call this the Refus al-as-W aiver Ac c ount.
The ac c ount is untenab le. Refusal-as-Waiver tells us that refusal has the same moral
eff e ct as c onsent . But it is clear that refusal has a v ery diff erent moral eff e ct. Unlike c onsent,
4
Se e, f or e x amp le, Jeff M cM ahan, “The Basis of M oral Liability to Def ensiv e Killing,” Phi losop hi ca l Issu es
15 (2005): 394-404; and Judith Jarvis Thomson, “Self-Def ense,” Phi losop hy a nd Pu b li c Aff a i rs 20 (1991): 302-
303.
44
the refusal of prote ction does not render it impermissib le f or the target to def end herse lf ,
nor does it strip her of her claims to re c eiv e c ompensation if the attack suc c e e ds. Imagine,
f or instanc e, a v ariant of Si b li ngs in which T eresa is not a p acifist, but in which she refuses
prote ction from Dan simp ly be c ause it matters to her that she be the one to def end herself and
that she be the one responsib le f or the harm to their brother Adam. In such a c ase, it se ems,
T eresa w ould be permitte d to def end herself even whi le Dan is not permitte d to def end her.
Refusal-as-Waiver pre dicts otherwise, ho w ev er, sinc e b y w aiving those of her rights that
are threatene d b y Adam, T eresa w ould fre e Adam of his liability to suff er def ensiv e harm at
the hands of everyone — herself include d.
Likewise, imagine that T eresa f ails to def end herself and that Adam’s attack suc c e e ds: he
breaks T eresa ’s leg. The f act that T eresa refuse d Dan’s prote ction surely does not change
the f act that Adam no w o w es T eresa c ompensation f or her injuries. Refusal-as-Waiver
pre dicts otherwise, ho w ev er. When a person w aiv es their right against some harm, they are
not typic ally o w e d c ompensation f or that harm. (Illustration: When I de cide to donate m y
kidney and c onsent to the doctor’s remo v al of the kidney , the doctor does not o w e me c om-
pensation f or m y loss.)
W e should reje ct Refusal-as-Waiver . And sinc e this ma y se em like the only game in
to wn f or e xp laining the po w er of refusal at the liability dimension, it ma y se em that w e should
theref ore reje ct the idea that refusal w orks its magic b y undermining attacker liability , and ac-
c ept that it must theref ore w ork its magic at the reasons dimension. In Se ction §2.4 I argue
that this is too quick, and that there are w a ys in which a target’s refusal c an undermine at-
tacker liability without c osting the target her rights. But bef ore w e get to m y positiv e ac c ount,
w e w ould do w ell to e x amine the only dev elope d attempt to ac c ount f or the po w er of refusal
45
at the reasons dimension.
2.3 The Exclusion Account
The ac c ount c omes c ourtes y of Jonathan P arry .
5
Ac c ording to P arry , among our normativ e
po w ers is the po w er to e x clude our w ell-being from c ontributing to another’s reasons:
P o w er to Ex c l ud e . A person X has the normativ e po w er to make it so that the
f act that Y’s φ -ing w ould benefit (or harm) X is not a reason f or Y (not) to φ .
6
On P arry’s view , the refusal of prote ction is just one w a y of e x ercising this po w er. Consider
Si b li ngs . Dan ma y def end T eresa if she does not refuse his prote ction. Absent her refusal,
Adam is liab le to be harme d b y Dan and there is sufficient reason to impose such harm (sinc e
harming Adam will greatly benefit T eresa). But T eresa has the po w er to e x clude her w ell-
being from c onsideration. That is, she has the po w er to make it so that ha rmi ng A da m wou ld
save T eresa from a bro ken leg is no longer a reason f or Dan to harm Adam. T eresa e x ercises
this po w er b y w a y of refusing Dan’s prote ction. Thus, when T eresa refuses prote ction, she
makes it so that the reasons c ondition on def ensiv e harm is no longer satisfie d f or Dan: Dan
no w lacks sufficient reason to harm Adam.
Let’s c all the c onjunction of P o w er to Ex clude with the claim that the refusal of prote ction
is one w a y of e x ercising this po w er the Ex cl usion Ac c ount. This ac c ount nic ely a v oids the
af orementione d shortc omings of Refusal-as-Waiver . Exclusion permits T eresa to de-
f end herself ev en as she refuses Dan’s prote ction. This is be c ause po w ers of reasons-e x clusion
are person-re lative : T eresa ’s refusal of Dan’s prote ction only makes it so that her benefit does
5
P arry , “Def ensiv e Harm,” 356-3 96.
6
P arry , “Def ensiv e Harm,” 371, c alls this the Power of Pru dentia l Exc lusi on princip le.
46
not c ount as a reason f or Da n to def end her; it does not likewise e x clude T eresa ’s w ell-being
from her poo l of reasons. Likewise, Exclusion also a v oids the imp lic ation that refusal un-
dermines the refuser’s right to c ompensation. Ac c ording to Refusal-as-Waiver , T eresa
w aiv es those of her rights that are threatene d b y Adam. This means that when Adam breaks
her leg, T eresa is harme d but her rights are not infringe d. And be c ause her rights are not in-
fringe d, she is not o w e d c ompensation. On Exclusion , b y c ontrast, T eresa does not giv e up
an y of her rights when she refuses prote ction. Thus, when Adam breaks her leg, her rights
a re infringe d. It is f or this reason that she is still o w e d c ompensation ev en though she refuse d
prote ction.
Despite these virtues, I argue, w e should reje ct Exclusion . M y obje ction c omes in thre e
p arts. First, I argue that the P o w er of Ex clusion princip le has v ery imp lausib le imp lic ations.
Se c ond, I argue that P arry’s positiv e argument f or P o w er of Ex clusion ac c ount f ails; the prin-
cip le is under-motiv ate d. Third, I argue that ev en if w e grant the truth of P o w er of Ex clusion,
w e should reje ct the idea that the refusal of prote ction is generally use d to e x ercise this po w er
of reasons-e x clusion.
2.3.1 Implausible Implications
Consider:
Ru nawa y Bou lder . An earthquake has sent a boulder c are ening do wn a hill to-
w ards Alic e. If it hits Alic e, it will kill her. Y ou c an let the boulder hit Alic e, or
y ou c an re dire ct it to the left or right. If y ou re dire ct it to the left it will hit and
kill Bob . If y ou re dire ct it to the right it will merely break one of Charlie’s toes.
Bob sinc erely tells y ou not to take his interests into ac c ount, and asks that y ou
47
sac rific e hi m rather than let Alic e or Charlie suff er harm.
If there is such a thing as a po w er of reasons-e x clusion, Bob clearly e x ercises it here. He
is v ery e xp licit about the ro le he w ants his interests to p la y in y our de cision. Giv en P o w er to
Ex clude, then, Bob’s e x ercise of this po w er makes it so that t he f act t hat Bo b wou ld be ki l led
does not c ount as a reason to not dire ct the boulder to w ard Bob . But this w ould mean that
y ou ha v e greater reason to re dire ct the boulder to w ard Bob than to w ard Charlie. It w ould
not only be permissib le to re dire ct the boulder to w ard Bob , but the thing that y ou s hou ld do .
I find this v ery imp lausib le. P erhaps it is true that y ou w ould not wrong Bob b y killing
him; he has, perhaps, c onsente d to this f ate. But ev en so , it is ob vious that y ou s hou ld re dire ct
the boulder to w ard Charlie rather than Bob . E v en if w e grant that Bob has c onsente d to be
kille d, y ou p lainly ha v e more reason to re dire ct the boulder to w ard Charlie.
No w it might be argue d that w e ha v e more reason to harm Charlie than Bob , not be c ause
of Bob’s interests, but be c ause of the interests of others.
7
It might be thought, f or instanc e,
that the reasons f or harming Charlie are grounde d in the interests that Bob’s friends and
f amily ha v e in Bob’s leg not being bro ken. But these c annot p lausib ly be the only reasons
against turning the boulder to w ard Bob . The f act that y ou should not kill Bob is not p lausib ly
c ontingent on the number of pe op le who c are about Bob . (As a test, imagine that there is
no one in the w orld who c ares about Bob , but man y pe op le who c are about Charlie. Still, it
se ems clear, y ou should not kill Bob .) But more importantly , it just se ems ob vious that the
strongest reason not to dire ct the boulder to w ard Bob is that it wou ld ki l l Bo b ! Just as it is
7
P arry off ers this sort of rep ly on p . 382 in response to a similar c ase. He mentions tw o reasons to prote ct
somebody who has refuse d prote ction that are not grounde d in the interests of the target: there are impersonal
reasons to prevent i n justi c e , and there are personal reasons to prevent t he attac ker from c om mitti ng a n i n justi c e .
But neither of these reasons are present in a c ase like Ru nawa y Bou lder where the threat is a natural one. There
is no attacker.
48
p lain t hat y ou should not dire ct the boulder to w ard Bob , it is p lain why y ou should not dire ct
the boulder to w ard Bob .
Here’s another w a y w e might try to resist the troub ling imp lic ation I’v e attribute d to Ex-
clusion . Think about c onsent. Not all c onsent is va li d c onsent. That is, not all c onsent
suc c e e ds in doing what c onsent is suppose d to do — namely , to e x ercise one’s po w er of rights-
w aiv er. Consent c an f ail to be v alid on v arious proc e dural grounds (e.g., f or reasons of inc om-
petenc e, c oercion, de c eption). It c an perhaps also f ail to be v alid on the substantiv e ground
that the things c onsente d to c onflicts with one’s se lf-reg a rdi ng duti es (e.g., y ou c an’t v alidly
c onsent to be torture d). But — the line of thought c ontinues — if our po w er of rights-w aiv er
is like this, wh y not our po w er of reasons-e x clusion? Wh y not think that Bob’s refusal is just
an e x amp le of i nva li d refusal — an instanc e where refusal f ails as an e x ercise of the po w er of
reasons-e x clusion on the grounds that such an e x ercise w ould c onflict with his self-regarding
duties?
8
M y rep ly: P erhaps this line of rep ly c an help Exclusion to resist c ases like Ru nawa y
Bou lder where there is a massiv e mismatch betw e en the sev erity of harm threatening one
person and the sev erity of harm threatening another person (e.g., death f or Bob and a bro ken
toe f or Charlie). I’m not sure. But I agre e that it’s least not ob vious that w e don’t ha v e self-
regarding duties to not make massiv e sac rific es f or a v ery small benefit f or some one else.
Consider, though, c ases where there isn’t an y pa i rwise mismatch like this, but only a mis-
match that results from interpersonal aggregation of harms. Here’s a v ariant of Ru nawa y
Bou lder :
Ru nawa y Gia nt Bou lder. A (v ery large!) boulder ro lls to w ards one-million pe o-
8
Thanks to V ictor T adros and Karam vir Chadha f or pressing this line of rep ly on behalf of Exclusion .
49
p le (the As). It will kill them all if y ou do nothing. Y ou c an re dire ct the boulder
right or left. If y ou re dire ct it to the right, it will break the legs of one-million
pe op le (the Bs). If y ou re dire ct it left, it will break the legs of one person, C.
E ach of the Bs tells y ou to ignore their interests and c onsents to ha v e their legs
bro ken.
Here I find it impossib le to resist the idea that each of the Bs va li d ly c onsents to ha v e their
legs bro ken. If y ou re dire ct the boulder their w a y , it se ems they are not wronge d, they ha v e no
standing to c omp lain, they are not o w e d c ompensation, etc. In an y c ase, I find it imp lausib le
to think that each of the Bs has a self-regarding duty to not w aiv e their right to be rescue d from
bro ken legs in order to sa v e some one else from the same f ate. But if the Bs do not vio late their
self-regarding duties b y c onsenting to suff er bro ken legs here, then it’s imp lausib le to think
they vio late those duties b y e x cluding the reason-giving f orc e of those harms.
Nonetheless, I find it v ery imp lausib le that y ou should re dire ct the boulder to w ards the
Bs. Y ou w ouldn’t wrong an y of the Bs b y breaking their legs; they’v e c onsente d to this f ate.
But ev en so , it’s ob vious that y ou s hou ld re dire ct the boulder to w ards C. E v en if w e grant that
each of the Bs has c onsente d to be kille d, y ou p lainly ha v e more reason to re dire ct the boulder
to w ard C. The b asic prob lem remains: Exclusion pre dicts the elimination of reasons that
se em not to be eliminate d.
2.3.2 Exclusion and Paternalism
No w perhaps w e c ould bite the bullet in c ases like Ru nawa y Bou lder (or Ru nawa y Gia nt
Bou lder ) if there w as an e xtremely c ompelling argument in f a v or of P o w er to Ex clude. But
there isn’t. Or at least, I’ll argue, P arry does not suc c e e d in giving us one.
50
P arry’s main line of support f or P o w er to Ex clude is that the princip le p la ys a c entral ro le
in e xp laining“p aternalism’s distinctiv e wrongness.”
9
It is largely unc ontro v ersial that when
it is wrong t o act p aternalistic ally to w ard some one, this is so be c ause the action w ould den y
the person’s autonom y or c ompetenc y in some w a y . The hard question is ho w to make this
idea pre cise. P arry suggests that the best pre cisific ation will appeal to something like P o w er to
Ex clude. When m y benefit does not justify y our interf ering with m y autonom y ev en f or m y
o wn sake, sa ys P arry , this is be c ause, “ As a result of e x ercising m y will, c ertain reasons — those
pertaining to m y good or w elf are — no longer c ontribute to justifying others ’ actions.”
10
But this is not p lausib le as a general diagnosis of the wrongfulness of wrongful p aternalism.
Consider a p aradigm c ase:
Cig a rettes : M y brother willingly smo kes on oc c asion, but only be c ause he en-
jo ys it, and not be c ause of an y addictiv e c ompulsion. And when he smo kes, he
only ev er smo kes outside, where he is t he only person who suff ers harm. I c are
about m y brother’s health, ho w ev er, and so I se c retly hide his cigarettes from
him whenev er I find them, letting him think that he absent-minde dly misp lac e d
them.
I benefit m y brother b y hiding his cigarettes. But m y p aternalistic action is nonetheless
(pro tanto) wrongful. Wh y? Ac c ording to the P o w er t o Ex clude diagnosis, the answ er is
that t he f act t hat hi di ng t he cig a rettes benefits m y brot her has be en e x clude d from m y reasons
f or hiding the cigarettes b y an e x ercise of m y brother’s will. That is wh y this f act does not
c ontribute to a justific ation f or hiding the cigarettes.
9
P arry , “Def ensiv e Harm,” 372.
10
Ibid., 373.
51
I find this diagnosis imp lausib le. M y hiding the cigarettes is wrongfully p aternalistic whether
or not m y brother has “ e x ercise d his will ” in an yw a y that c an be said to be an e x ercise of his
po w er of reasons-e x clusion. He c ertainly c ou ld ha v e perf orme d c ertain actions whereb y he
e x ercise d this po w er. P erhaps he w ould e x ercise this po w er if he to ld me that he didn’t w ant
me to prev ent him from smo king. But the point is that he ne e dn’t do an y such thing f or m y
action to be wrongful. He doesn’t ne e d to perf orm some action that c an be p lausib ly be said
to e x clude c ertain f acts about his w ell-being from m y realm of reasons in order f or it to be
impermissib le f or me to hide the cigarettes. F orms of p aternalism like this one are wrongful
“b y def ault”; no e x ercise of po w ers of reasons-e x clusion are ne e de d to make them wrongful.
The P o w er to Ex clude diagnosis is ev en less p lausib le when app lie d to wrongful instanc es
of p aternalistic “nudging.”
11
I merely nudge some one to φ when,instead of c oercing them
to φ or making it impossib le or more difficult f or them to do otherwise, I motiv ate them to
φ b y altering their perc eption of their choic es. Not all nudg es are wrongful or p aternalistic.
But some are. F or e x amp le:
A dvi c e : An undergraduate student asks her p hilosop h y prof essor where she might
app ly to graduate schoo l, giv en that she’s intereste d in programs that are strong
in ethics but also intereste d in programs that are strong in metap h ysics. Sup-
pose the prof essor someho w kno ws that it w ould be better f or the student if
she w ere to study ethics than metap h ysics. And suppose the prof essor (but not
the student) also kno ws that this student has the f o llo wing ps y cho logic al trait:
if she hears some one snap their fingers while presenting an option, she will f orm
a strong, irrational pref erenc e to w ard that option. In order to benefit the stu-
11
The c onc ept of a nudge w as introduc e d b y Richard Thaler and Cass Sustein, “Libertarian P aternalism,”
The A meri ca n Ec onomi c Revi ew 93 (2003): 175–79.
52
dent, the prof essor e xp loits this trait, and subtly snaps his fingers only while
mentioning t he ethics programs.
Intuitiv ely , this “nudge” is wrongful and p aternalistic. But this is not be c ause the student
has done an ything to e x clude her w ell-being from the prof essor’s realm of reasons. On the
c ontrary: w e might suppose that she wa nts the prof essor to take her w ell-being into c onsid-
eration when giving her a list of schoo ls! The prob lem is not t hat her w ell-being is taken into
c onsideration, but the w a y into which it is taken into c onsideration. The prob lem is that the
prof essor takes her de cision out of the domain of her rational deliberations, and in so doing
treats her as lacking the c ap acity to make her o wn de cision about which field of p hilosop h y
to pursue.
The lesson: w e c annot giv e a general e xp lanation of the wrongfulness of wrongful p ater-
nalism b y appeal to P o w er to Ex clude. And thus w e c annot motiv ate this princip le on the
grounds that it p la ys a c entral ro le in pro viding a general e xp lanation f or p aternalism’s dis-
tinctiv e wrongness.
2.3.3 Refusal and the Exclusion of Reasons
I’v e thus f ar argue d that w e should reje ct P o w er to Ex clude. But ev en if this princip le w ere true
— ev en if w e ha v e such po w ers of reasons-e x clusion — w e should still reje ct the Exclusion
ac c ount. This is be c ause w e should reje ct the other c onjunct of this ac c ount: w e should reje ct
the idea that w e generally e x ercise this po w er of reasons-e x clusion by wa y of refusi ng someone ’s
protecti on . Consider:
Easy Defense : Attacker attempts to murder T arget. Def ender c an prote ct T ar-
get b y lightly punching Attacker. But T arget is such a strict p acifist that she
53
c ondemns ev en a mild harm such as this, and f or this reason refuses Def ender’s
prote ction
I ha v e a v ery diff erent intuitiv e reaction to c ases like this than to c ases like Si b li ngs . When
the prote ctiv e harm that c an be impose d on an attacker is mu c h less than the harm he threatens
on his target, I do not find it p lausib le that a w ould-be def ender is morally re quire d to refrain
from prote ctiv e interv ention. On the c ontrary , I find it intuitiv ely c ompelling that Def ender
s hou ld interv ene in such a c ase.
Exclusion , ho w ev er, deliv ers the same v erdict in Easy Defense as in Si b li ngs . Just like
T eresa, T arget has refuse d prote ction. She has thereb y e x clude d her w ell-being from De-
f ender’s reasons f or prote cting her. Def ender thus has no reason to prote ct T arget (ap art
from reasons that are grounde d in the interests of others), and he has some reason to not
prote ct T arget — namely , that this w ould c ause mild harm to Attacker.
12
Exclusion thus imp lies that Def ender lacks the permission to interv ene. I find this result
e xtremely c ounterintuitiv e. E v en if there is such a po w er as a po w er of reasons-e x clusion,
T arget is not p lausib ly e x ercising that po w er b y refusing prote ction.
12
One might think that Def ender has the f o llo wing “impersonal ” reason to prote ct T arget: that it will prev ent
T arget from suff ering (and Attacker from c ommitting) an injustic e (P arry , “Def ensiv e Harm,” 382). But such
reasons p lainly do not make the diff erenc e in Easy Defense , as w e c an se e b y c onsidering a v ariant of the c ase
in which the threat is purely natural: A storm has b lown a bou lder down a hi l l a nd towa rd T a rget; Defender
ca n protect T a rget by redi recti ng t he bou lder su c h t hat it wi l l mi ld ly bruise Bysta nder . As no one has pushe d the
boulder, T arget f ac es a threat of harm but not a threat of injustic e. But this doesn’t se em to change the intuitiv e
v erdict: still it is wildly c ounterintuitiv e that T arget c an, b y refusing prote ction, re quire Def ender to sit idly b y
while T arget is kille d. Def ender se ems p lainly permitte d to re dire ct the boulder. (I e xp lore the interp la y betw e en
refusal and lesser-evil justific ations in greater length in §2.5.)
54
2.4 The Liability Contraction Accoun t
The Exclusion ac c ount claims that the po w er of refusal is loc ate d at what I’v e c alle d the
“reasons dimension.” Spe cific ally , it claims that refusal undermines the permissibility of pro-
te ctiv e interv ention b y e x cluding c ertain reasons that w ould normally f a v or such interv en-
tion. I’v e argue d that w e shou ld reje ct this ac c ount. But t his isn’t the only w a y that refusal
might make a diff erenc e at the reasons dimension. Rather than e x cluding some of one’s rea-
sons for prote ctiv e interv ention, refusal might instead c ontribute additional reasons ag a i nst
prote ctiv e interv ention.
Refusal surely does c ontribute such reasons. Consider our c entral c ase of Si b li ngs . When
T eresa refuses Dan’s prote ction, T eresa rev eals c ertain of her pref erenc es to Dan. Sinc e it is
(pro tanto) in T eresa ’s interest that her rev eale d pref erenc es are satisfie d, this c onstitutes one
reason f or Dan to refrain from prote cting her. Likewise, it ma y be (pro tanto) in T eresa ’s
interest that others e xhibit respe ct f or her de ep ly held v alues and c ommitments, such as her
p acifism. This too w ould giv e Dan some reason to refrain from prote cting her.
But while the refusal of prote ction will almost alw a ys giv e the w ould-be def ender some
reasons to refrain from interv ention, I think it is rare that such reasons are w eighty enough
to tip the sc ale. It is in T eresa ’s interest that her rev eale d pref erenc es are satisfie d and that
others e xhibit respe ct f or her v alues and c ommitments. Still, her interest in not being seri-
ously harme d greatly outw eighs these interests. This is espe cially clear in a v ersion of Si b li ngs
where Adam will ki l l T eresa unless Dan interv enes. E v en if w e grant that T eresa has an in-
terest in pref erenc e-satisf action and respe ct f or her p acifistic c ommitments, it is strongly in
T eresa ’s a l l-t hi ngs-c onsi dered interest that Dan interv ene. So while refusal c ertainly giv es the
55
refuser more of an interest in the def ender’s refraining from interv ention, this is not usually
sufficient on its o wn to render otherwise-permissib le prote ction impermissib le. There must
be m ore to the story .
I think w e must loo k bey ond the mere i nterests at stake and c onsider the possibility that
refusal thro ws additional rig hts -b ase d obstacles into the mix. Rights ha v e the sort of profile
w e ne e d: they are the sort of things that impose v ery w eighty c onstraints on action - c on-
straints that c an outw eigh ev en the sa ving of a person’s lif e. (I am not permitte d, e.g., to kill
one person so as to harv est his organs to sa v e the liv es of tw o others.) If refusal c ontributes
a right such that prote ctiv e interv ention w ould infringe that right, this w ould nic ely e xp lain
wh y refusal often undermines the permissibility of such prote ction.
In this se ction, I articulate one v ersion of such an ac c ount. T o introduc e the ac c ount w e
should start b y noticing an important f eature of our rights against harm: when these rights
are threatene d b y an attacker, some of the attacker’s o wn rights against harm are thereb y lost.
Spe cific ally , the attacker c omes to lose those rights that “stand in the w a y” of the def ense
of y our right: he loses his rights against (ne c essary and proportionate)harm that is a means
or side-eff e ct of prev enting him from infringing y our right.
13
Importantly , an attacker does
not just lose such rights with respe ct to his target; he loses these rights with respe ct to an y
potential def ender. When Adam attacks T eresa, f or instanc e, he loses his right that T eresa
13
Our rights against harm ha v e this f eature, I think, be c ause our interests a re better serv e d b y a regime of
rights with this f eature than a regime without it. Comp are tw o moral regimes. In Regime
1
, when X threatens
Y’s rights, X loses c ertain rights — namely rights against suff ering ne c essary and proportionate def ensiv e harm.
In Regime
2
X does not lose such rights. Regime
1
off ers better prote ction f or the ta rgets of rights-infringements,
sinc e it remo v es a moral b arrier to the def ense of such persons. Regime
2
off ers better prote ction f or t hose who
t hreaten the rights of others, sinc e they enjo y a moral b arrier against suff ering def ensiv e harm. But the w a y these
adv antages are distribute d in Regime
1
se ems p lainly more just than in Regime
2
. This is be c ause the c omp arativ e
prote ctiv e disadv antages a person suff ers in Regime
1
are the product of their o wn choic es and actions, whereas
the prote ctiv e disadv antages a person suff ers in Regime
2
are the products of some one else’s choic es and actions.
56
not harm him as a means or side-eff e ct of a v erting his attack — but so too he loses his right
that Da n not harm him as a means or side-eff e ct of a v erting his attack. Adam makes himself
liab le from all sides, so to speak.
This is the def ault eff e ct of posing a threat to another’s rights against harm. E arlier w e
note d an unc ontro v ersial w a y in which it is in a target’s po w er to mitig ate these eff e cts: she
c an w aiv e the right that is threatene d. By w aiving that right, she thereb y restores all of the
rights that had be en lost b y the attacker. He regains those rights he had lost with respe ct to
his target as w ell as those rights he had lost with respe ct to other potential def enders.
It matters to us that w e ha v e c ontro l o v er our rights, and the po w er to w aiv e our rights giv es
us a f orm of c ontro l o v er our rights w e w ould not otherwise ha v e. It lets us c ontro l whether
or not the normal prote ctions of our rights remain in eff e ct. But notic e: the po w er to w aiv e
our rights is a v ery b lunt too l b y which to e x ercise such c ontro l. When T w aiv es a right that
A n ot φ , she eliminate s all of the prote ctions aff orde d b y that right:
• The w eighty obstacle that this right presents to the permissibility of A ’s φ -ing is re-
mo v e d. ( Deonti c prote ction)
• A regains his v arious rights not to be harme d as a means or side-eff e ct of prev enting his
φ -ing. ( Lia bi lity prote ction)
• The f o llo wing c onditional is no longer t rue: if A suc c e e ds in φ -ing, T has a right that
A p a y c ompensation f or her loss. ( Compensatory prote ction)
The po w er to w aiv e our rights off ers us a f orm of c ontro l o v er our rights, but it is a v ery
all-or-nothing f orm of c ontro l. W e enjo y better c ontro l o v er our rights, ho w ev er, if w e also
57
enjo y more fine-graine d po w ers with which to mitigate their prote ctions.
14
If c onsiderations
of c ontro l justify admitting into our normativ e the ory the po w er to remo v e the c omp lete
p ackage of prote ctions aff orde d b y rights, these same c onsiderations should justify admit-
ting in the po w er to remo v e only p arts of the p ackage.And it often matters to us that w e ha v e
this more fine-graine d f orm of c ontro l. Consider, f or instanc e, the c ase of Si b li ngs . T eresa
is a p acifist, who does not w ant others to def end her and who does not w ant to def end her-
self either. Still, she does not w ant to be harme d, and she w ants to be c ompensate d if she is
harme d. Rather than w aiv e her right in its entirety , T eresa w ould pref er the po w er to remo v e
the def ault permission that others ha v e to prote ct her without also giving up her de ontic and
c ompensatory prote ctions. Or to take a diff erent c ase: imagine a prof essional mix e d martial
artist is attacke d on the stre et; an onloo ker c an def end her, but the fighter c an just as w ell de-
f end herself with the same amount of def ensiv e f orc e. And suppose it matters to the fighter
that her lif e be characterize d b y self-sufficienc y and that she be the one to def end herself. The
fighter w ould not w ant to w aiv e her rights in this c ase, sinc e she w ouldn’t w ant to giv e up her
permission to def end herself. She w ould only w ant the po w er to giv e up the permission that
ot her pe op le ha v e to d ef end her.
I propose that w e ha v e just such a po w er to remo v e some — and only some — of the
lia bi lity prote ctions of our rights without also remo ving the de ontic and c ompensatory pro-
te ctions of our rights. Spe cific ally , w e ha v e the po w er to restrict the domain of persons with
respe ct to whom our attacker is liab le. Where the normal eff e ct of an attack is to make the at-
tacker liab le to an y potential def ender, this eff e ct c an be mitigate d b y remo ving the attacker’s
14
There is some degre e of sele ctivity that c an be e x ercise d with the po w er of w aiv er. F or e x amp le, I c an w aiv e
m y c ontingent right to re c eiv e c ompensation if y ou break m y arm without w aiving m y right that y ou not break
m y arm. But m y point abo v e is that there are man y w a ys w e might w ant to sele ctiv ely mitigate the eff e cts of our
rights that w e c annot ac c omp lish b y simp ly w aiving a right. W aiv er off ers some sele ctivity , but not enough.
58
liability with respe ct to on ly some potential def enders. It is this v ery po w er, I claim, that is typ-
ic ally e x ercise d b y the act of the refusal of prote ction, and that e xp lains the bulk of refusal ’s
moral po w er. When T eresa refuses Dan’s prote ction, f or instanc e, she does not w aiv e her
right. But she does mitigate one of the normal eff e cts of the threat to her rights: she makes
it so that Adam is no longer liab le to be harme d by Da n . This is wh y Dan lacks a permission
to def end T eresa. And be c ause T eresa on ly remo v es Adam’s liability with respe ct to Dan,
Adam remains liab le to be harme d b y T eresa. This is wh y T eresa — unlike if she had simp ly
w aiv e d her right — retains her right against Adam’s attack, retains her permission to def end
herself, and retains her c onditional right to c ompensation if the attack suc c e e ds. Call this the
Liability Contra ction Ac c ount (“ Contraction ” f or short).
2.5 Features and Implications
The b asic idea of Contraction to hand, let’s unp ack some of the imp lic ations and distinc-
tiv es of the ac c ount.
2.5.1 Wronging the Attacker
Suppose Dan disregards T eresa ’s demand that he not prote ct her. He breaks Adam’s leg,
thereb y prev enting Adam from breaking T eresa ’s leg. Contraction tells us that Dan acts
impermissib ly , sinc e Adam is not liab le to ha v e his leg bro ken by Da n . Contraction , then,
imp lies that Dan vio lates Adam’s rights. This imp lic ation ma y giv e man y readers p ause. If
Dan wrongs an y one, isn’t it T eresa rather than Adam?
Notic e, first, ho w ev er, that Contraction does not imp ly that T eresa is no t a lso wronge d
b y Dan. On the c ontrary , the ac c ount off ers resourc es to e xp lain wh y T eresa is wronge d. I
59
think it p lausib le that w e ha v e a right that others respe ct our normativ e po w ers — a right
that they not disregard the eff e cts of those po w ers and act as though w e lack such po w ers.
Consider, f or e x amp le, a c ase in which some one disregards another’s po w er of w aiv er:
Slas her : Jill v alidly c onsents to ha v e se x with Jack later that night. Bloggs kno ws
what Jill has said to Jack, but due to c ertain se xist assumptions has the belief that
Jill lacks the po w er to w aiv e her right not to be intimately touche d b y Jack. T o
“prote ct” Jill, Bloggs slashes the tires on Jack’s c ar so that he c an’t driv e to Jill ’s.
Bloggs does not just wrong Jack; he also wrongs Jill. And he wrongs Jill ev en if her interests
are not set b ack—ev en if, f or e x amp le, she w ould be just as happ y (or happier) to read her
no v el that night as she w ould to ha v e se x with Jack. Jill is wronge d ev en if she is not harme d
pre cisely be c ause Bloggs f ails to treat her as ha ving a normativ e po w er that she in f act has. He
f ails to respe ct her agenc y in this w a y .
Giv en Contraction , Dan w ould wrong T eresa in this same w a y b y disregarding her
demand that he not prote ct her: Dan w ould treat T eresa as lacking po w ers she has to mitigate
some of her o wn rights-b ase d prote ctions. Contraction , then, nic ely ac c ommodates the
intuition that Dan’s interv ention w ould infringe T eresa ’s rights.
But what should w e sa y about the imp lic ation that Dan also infringes A da m’s rights? This
will se em c ounterintuitiv e to man y , in light of the further imp lic ations this ma y se em to ha v e.
It ma y se em to f o llo w from the f act that X infringes Y’s rights that (i) Y has the standing to
c omp lain about the infringement, (ii) Y w ould ha v e permission to def end himself from the
infringement, and (iii) X o w es Y c ompensation f or the infringement. But it se ems imp lausi-
b le that an y of these things are true with respe ct to Adam and Dan. If Dan disregards T eresa ’s
demand and interv enes to prote ct her, it se ems that Adam has no standing to c omp lain, no
60
permission to use c ounter-def ensiv e f orc e against Dan, and no right to c ompensation if Dan
breaks his leg.
In f act, ho w ev er, this is c omp atib le with the claim that Dan infringes Adam’s rights. First,
it does not f o llo w from the f act that X infringes Y’s right that Y has the standing to c omp lain
about the infringement. Suppose I regularly break promises that I make to others. F or this
reason, I will lack the standing to c omp lain if y ou break a promise to me: to c omp lain w ould
be to ho ld y ou to standards to which I do not subje ct m yself. Nonetheless, this does not
mean that y ou f ail to infring e m y rights b y breaking y our promise. The f act that I regularly
break promises to others does not make it impossib le f or y ou to make binding promises to
me — to make promises that impose on y ou a dire cte d duty (and on me a c orrelativ e right)
that y ou do the thing y ou promise d to do .
I think the v ery same c onsiderations are true of Adam. E v en on the assumption that Dan
infringes Adam’s rights, w e should think that Adam lacks the standing to c omp lain. He lacks
the standing to c omp lain that his leg is bro ken f or the v ery same reasons I lack the standing
to c omp lain about y our bro ken promise: to c omp lain in these circumstanc es w ould bef ore
Adam to ho ld Dan to standards (“ don’t break some one’s leg without justific ation”) to which
he does not subje ct himself.
Se c ond, it does not f o llo w from the f act that X infringes Y’s right that Y w ould be permit-
te d to use (proportionate) c ounter-def ensiv e f orc e nor that Y is o w e d c ompensation f or the
infringement. T o se e wh y , start b y c onsidering:
Mi ners : Thre e miners — Alp ha, Beta, and Gamma — are de ep in a mine when
they enc ounter a harmful gas. If e xpose d to the gas f or an e xtende d period,
they will suff er moderate lung damage. The thre e miners quickly put on their
61
gas masks. Alp ha, ho w ev er, is jealous of Beta ’s re c ent promotion, and attacks
him, destro ying his gas mask. Beta no w must choose betw e en suff ering lung
damage or stealing Gamma ’s mask (it is impossib le f or him to take Alp ha ’s mask
b y f orc e). He de cides to attack Gamma and steal her mask. No w it is Alp ha who
f ac es a choic e: he c an def end Gamma either b y knocking Beta to the ground (in
which c ase Beta will suff er the lung damage) or b y giving up his o wn gas mask
to Beta (i n which c ase Alp ha will suff er the lung damage).
Beta poses a wrongful threat to Gamma. And y et it is intuitiv ely c ompelling that Alp ha is
not permitte d to knock Beta to the ground in def ense of Gamma. Alp ha is not permitte d to
knock Beta to the ground be c ause Alp ha has a duty to assume the c osts of his o wn wrong-
doing. This is wh y he must giv e up his mask rather than knock Beta to the ground or allo w
Beta to take Gamma ’s mask: giving up his mask shifts the c ost a w a y from Beta or Gamma
and onto himself. Alp ha ’s duty to assume these c osts thus functions to b lock his permission
to impose prote ctiv e harm on Beta.
I think Adam is f orbidden from def ending himself against Dan in Si b li ngs f or the same
reasons that Alp ha is f orbidden from imposing def ensiv e harm on Beta in Mi ners . Alp ha has
a duty to assume the c osts of Beta ’s attack on Gamma be c ause of f acts about his responsi bi lity
f or the c onditions of that attack. Alp ha is responsib le f or breaking Beta ’s gas mask. And b y
breaking Beta ’s mask, he puts B eta in a position in which Beta c an act permissib ly (i. e., not
attack Gamma) only at great c ost (i. e., damage to his lungs). This f act greatly mitigates Beta ’s
o wn responsibility f or the threat that he poses to Gamma. This diff erenc e in responsibility
betw e en Alp ha and Beta, I believ e, is what e xp lains wh y Alp ha has a duty to assume the c osts
of Beta ’s threat to Gamma (i. e., wh y he is not allo w e d to let those c osts f all on Gamma or to
62
re dire ct them onto Beta).
W e get a similar diff erenc e betw e en Adam and Dan. Adam is highly responsib le f or his
threat to T eresa. By posing this threat, Adam puts Dan in a position in which the latter c an
act permissib ly (i. e., refrain from prote cting T eresa) only at great c ost (i. e., a bro ken leg f or
his sister). This f act, I think, mitigates some of Dan’s responsibility f or his wrongful threat to
Adam. Giv en t his diff erenc e in c omp arativ e responsibility betw e en Adam and Dan, Adam
has a duty to assume the c osts of Dan’s attack. He isn’t liab le to be ta rgeted . But onc e he has
be en targete d, Adam t hen takes on a duty to assume the harm. And here this duty doesn’t
re quire Adam to activ ely re dire ct an y sort of c osts to w ard himself (as in the Mi ners c ase); it
re quires him only to abstain from re dire cting the threat b ack to w ard Dan. This is wh y Adam
is not permitte d to def end himself from Dan’s wrongful threat. This is also wh y Adam is not
o w e d c ompensation when his leg is bro ken. The reasons wh y Adam has an ex a nte duty to
assume the c osts of Dan’s attack — namely , the f act that it is more just if the c osts f all on the
p arty most responsib le f or those c osts — are the same reasons wh y Dan lacks an ex post duty
to assume the c osts of Dan’s attack.
It’s important to notic e, ho w ev er, that these c onsiderations do not generalize to every c ase
of refuse d prote ction. This is be c ause the attacker w on’t alw a ys bear a greater share of respon-
sibility than the def ender. Suppose, f or e x amp le, that Adam’s responsibility is also mitigate d:
Adam is mentally imp aire d, a villain has f e d him lo ads of f alse inf ormation, and the villain has
promise d to seriously harm his mother if Adam does not attack T eresa. I think that under
these c onditions Adam w ould be no more responsib le than Dan f or the c osts that must no w
be distribute d. He w ould not ha v e a duty to assume the c osts of Dan’s threat. Giv en Con-
63
traction , then, Adam wou ld be permitte d to def end himsel f from Dan in this c ase.
15
Likewise, c onsider:
Mista ken Attac ker : The identic al twin brother of a notorious murderer is driv-
ing late at night in a remote, English-speaking region when his c ar breaks do wn.
Una w are that his brother has re c ently esc ape d from prison and is kno wn to be
hiding in this same area, T win knocks on the door of the nearest house, se eking
to p hone f or help . The resident opens the door. T win, ho w ev er, speaks little En-
glish, and so does his best to c ommunic ate with frantic gestures. The frightene d
resident, ho w ev er, f orms the reasonab le belief that the harmless twin is the mur-
derer, and that his motions are vio lent threats. Resident lunges at T win with a
knif e. A neighbor who speaks T win’s nativ e language observ es the mistake, and
pulls out his o wn knif e with which to def end T win. But T win, a c ommitte d
p acifist, tells Neighbor not to interv ene. Neighbor attacks Resident an yw a ys.
16
Here again w e find a c ase in which a wrongful attacker’s responsibility is mitigate d. In this
15
I ho ld a similar view about c ounter-def ense against u n nec essa ry harm. An attacker attempts to kill y ou; y ou
c an e qually w ell def end y ourself b y killing him or b y disarming him with a light punch. Suppose y ou kno w this,
but still choose to kill the attacker rather than disarm him. I do not think the attacker is liab le to be kille d in
such circumstanc es. Does this mean the attacker c an fight b ack? It depends: whether and to what e xtent the
attacker c an use c ounter-def ensiv e f orc e depends (in p art) on f acts about c omp arativ e responsibility . Be c ause
attackers tend to be the “first c ause” and be c ause def enders often act under c onditions of limite d inf ormation
and duress in choosing ho w to def end themselv es, attackers will often bear a greater share of responsibility than
the def ender f or the f act that there is unne c essary harm they must no w choose to distribute. When this is the
c ase, the attacker will ha v e a duty to assume the c osts of the def ender’s o v erreaction, ev en if the def ender is lia b le
to c ounter-def ensiv e harm. But it isn’t alw a ys the c ase that an attacker is more responsib le f or his f orc e d choic e
than the def ender. Suppose y our attacker attempts to kill y ou only be c ause a villain will kill his f amily if he
doesn’t. And suppose y ou kno w y ou c an easily def end y ourself b y lightly punching the attacker, but a ttempt to
kill the attacker an yw a ys. Y ou ma y then be more responsib le than the attacker f or the f act that he must choose
betw e en letting himself be kille d and killing y ou. In such a c ase, he w ould lack a duty to assume the c osts of
y our def ensiv e o v erreaction and be permitte d to fight b ack. This point is discusse d in more detail in chapter
thre e (§3.5.2).
16
This c ase is adapte d from M cM ahan, “The Basis of M oral Liability ,” 387.
64
instanc e, Resident’s responsibility is mitigate d b y f acts about his epistemic situation. He is
justifie d in attacking T win in the evidenc e-relativ e sense. This doesn’t make him immune to
def ensiv e harm; he is still liab le. But it does, I think, mitigate his responsibility to such an e x-
tent that he lacks a duty to assume the c osts of Neighbor’s impermissib le interv ention. And
thus, giv en Contraction , he is permitte d to def end himself against Neighbor’s impermis-
sib le interv ention and w ould be o w e d c ompensation if Neighbor’s attack suc c e e ds.
M a y wrongful attackers def end themselv es against impermissib le prote ctiv e interv ention?
The answ er that Contraction giv es us is: someti mes . When an attacker bears a signific ant
amount of responsibility f or his threat, he will typic ally be f orbidden from def ending him-
self ev en from wrongful interv ention. But when an attacker’s responsibility is signific antly
mitigate d, he will sometimes be permitte d to def end himself.
2.5.2 Group Refusal
Another important imp lic ation of Contraction c onc erns the eff e cts of refusal in group
c ases. Consider, f or e x amp le:
Ma n y T a rgets : Attacker prep ares to thro w a grenade into a room with man y
pe op le. If Attacker thro ws the grenade ev ery one in the room will lose a hand.
He c an be stoppe d only if Def ender kills Attacker first. But some members of
the targe t group refuse Def ender’s prote ction.
Is Def ender permitte d to interv ene? If w e stipulate that every member of the target group
refuses Def ender’s help , then it se ems he is not permitte d to interv ene. But suppose, instead,
that — as often happens in real-w orld c ases of prote ctiv e, military interv ention — the group
65
is not of one mind. Some members of the target group refuse prote ction; some w elc ome it.
What then?
One might be tempte d to think that what matters in group c ases is the proportion of the
group that refuses prote ction. That is, one might be tempte d to think there is some fix e d per-
c entage, p %, such that if more than p % of the group refuses prote ction, otherwise-permissib le
prote ctiv e interv ention is rendere d impermissib le. M an y the orists ha v e endorse d this ap-
pro ach,
17
but Contraction joins P arry’s Exclusion Ac c ount in reje cting it. Both ac-
c ounts imp ly that there is no such fix e d perc entage. In some c onte xts, refusal b y a v ery small
perc entage of the group will render prote ctiv e interv ention impermissib le. But in other c on-
te xts, only a larger perc entage (sometimes unanimity) will do the job .
W e c an se e wh y this is when w e c omp are the f o llo wing tw o v ariants of the Ma n y T a rgets
c ase:
Ha nds 1 : Attacker prep ares to thro w a grenade into a room with one-thousand
pe op le. If he is not stoppe d, ev ery one in the room will lose a hand. Attacker
c an be stopp e d only if Def ender kills him.
Ha nds 2 : As abo v e, e x c ept that Attacker thro ws the grenade into a room with
only thre e pe op le.
Let’s start with Exclusion . F or the sake of illustration let’s stipulate that there is only
sufficient reason to ki l l an attacker when doing so will sa v e t hree or more peop le from the
17
By f ar the most popular position is one ac c ording to which only u na ni mous refusal c an render otherwise-
permissib le prote ction of a group impermissib le. Se e Altman and W ellman,“From Humanitarian Interv ention
to Assassination,” 242–45; F abre, “Cosmopo litanism,” 972–74; F abre, “P ermissib le Rescue Killings,” 159n11;
Fro w e, “JudgingArme d Humanitarian,” 108–09; M cM ahan, “Humanitarian Interv ention,” 51–53. But a ma-
joritarian princip le is a natural alternativ e, which might be def ende d on democ ratic grounds. Se e Kalle Grill,
“Liberalism, Altruism, and Group Consent,” Pu b li c Hea lt h Et hi cs 2 (2009): 146–57; and P arry , “Def ensiv e
Harm,” 367–68.
66
loss of a hand. Giv en this stipulation, Exclusion tells us that it is permissib le to thro w the
grenade if and only if at least thre e members of the target group do not refuse prote ction.
This is be c ause the interests of those who refuse prote ction do not c ontribute to Def ender’s
reasons to thro w the grenade. And this is true no matter ho w large the target group . Thus,
w e get the result that in Ha nds 1 prote ction is rendere d impermissib le only if at least 99.8%
of the target group refuses prote ction, but in Ha nds 2 prote ction is rendere d impermissib le
b y the refusal of a mere 33%. The important point is just that the proporti on of refusers to
non-refusers doesn’t matter to the permissibility of prote ctiv e interv ention. What matters is
the nu m ber of p ersons who do not refuse prote ction.
Contraction likewise tells us to ignore the perc entages, at least when w e c onjoin the ac-
c ount with m y pref erre d w a y of understanding narro w proportionality (as def ende d in chap-
ter o ne):
Pr opor tion ality as Demandingness . Def ensiv e harm impose d on an
attacker is narro wly proportionate just in c ase the sev erity of that harm does not
e x c e e d the aggregate dema ndi ngness of the attacker’s duties to a v ert the threats
that are a v erte d b y t hat def ensiv e harm.
18
Dema ndi ngness is a f eature of a duty distinct from its end. A duty’s end is its obje ctiv e, its
go al — the state of aff airs the agent has a duty to se e to . The demandingness of a duty is the
c ost at which the duty-bearer must pursue that end. M ore pre cisely , the demandingness of
a duty is the upper limit of c ost the agent is re quire d to bear in servic e of that duty: it is the
amount of c ost such that an y greater c ost w ould make it permissib le f or the agent to f ail to
discharge the duty . What it is narro wly proportionate to impose b y w a y of def ensiv e harm, I
18
This princip le is a generalize d v ersion of what I c alle d “Demands of Proportionality” in chapter one.
67
claim, just is the amount of c ost w e c ould ask the attacker to bear in the discharge of his duty
to a v ert his threat.
19
It is the amount of c ost he c ould be aske d to assume to a v ert his o wn
attack w ere he ab le to do so .
The demandingness of duti es c an aggregate. Suppose I c ould be aske d to bear up to c ost c
1
to se e to it that I not break A ’s toe. And suppose I c ould be aske d to bear up to c ost c
2
to se e
to it that I not break B’s toe. Then, all else e qual, I c ould be aske d to bear up to c ost c
1
+ c
2
to se e to it that I not break the toes of both A and B.
Here’s what this means f or Ha nds 1 and Ha nds 2 . F or the sake of illustration let’s stipulate
that lethal f orc e is a narro wly proportionate response to a threat of thre e — but not less than
thre e — sev ere d hands. Giv en Proportionality as Demandingness, this stipulation amounts
to the f o llo wing: the aggregate demandingness of Attacker’s duties to se e to it that he not
sev er two hands is not so great as to re quire him to lose his lif e in servic e of those duties,
but the aggregate demandingness of his duties to se e to it that he not sev er t hree hands is
great enough to re quire him to lose his lif e in servic e of those duties. Giv en Contraction ,
then, Attacker is liab le to be kille d b y Def ender only if at least thre e of his duties-not-to-
harm ground a liability with respe ct to Def ender. When one of the victims refuses Def ender’s
prote ction, ho w ev er, she makes it so that Attacker’s duty to w ards her no longer grounds
a liability to harm from Defender , thereb y remo ving its c ontribution to the upper limit of
harm Def ender c an impose on Attacker. F or this reason, prote ctiv e interv ention is rendere d
narro wly disproportionate in both Ha nds c ases only when the number of non-refusers drops
belo w thre e. This is true no matter ho w large the target group .
19
Se e §1.3.2.
68
2.5.3 Lesser-evil Justifications
The tw o the ories, ho w ev er, disagre e about wh y less-than-unanimous refusal renders prote c-
tion impermissib le when it does. Exclusion tells us that it is b y rendering prote ction wi de ly
disproportionate. Contraction tells us that it is rather b y rendering prote ction na rrowly
disproportionate.
20
M ore pre cisely , the tw o ac c ounts endorse diff erent instanc es of the f o l-
lo wing schema: Protective ha rm is [proporti onate] to defend ag a i nst t he t hreat to a grou p on ly if
t hat sa me ha rm wou ld be [proporti onate] to defend ag a i nst on ly t he t hreat to t he non-r efusi ng
su bset of t he grou p . Exclusion endorses the instanc e in which “[proportionate]” is rep lac e d
with “widely proportionate.” Contraction endorses the instanc e in which “[proportion-
ate]” is rep lac e d with “narro wly proportionate”.
This diff erenc e has important upshots f or the a v ailability of lesser-evil justific ations. First,
lesser-evil justific ations (as w ell as “mix e d ” or “ c ombine d ” justific ations) are easier to c ome b y
on Contraction than Exclusion .
21
T o illustrate, re c all the c ase of:
Easy Defense : Attacker attempts to murder T arget. Def ender c an prote ct T ar-
get b y lightly punching Attacker. But T arget is such a strict p acifist that she
c ondemns ev en a mild harm such as this, and f or this reason refuses Def ender’s
20
This c ommon termino logy originates with M cM ahan. Ac c ording to M cM ahan, there are tw o diff erent
proportionality c onstraints on def ensiv e harm. The na rrow proportionality c onstraint f ocuses on the attacker;
it re quires that the def ensiv e harm impose d on the attacker not be “too great” relativ e to the attacker’s threat.
The wi de proportionality c onstraint, b y c ontrast, f ocuses not (just) on the attacker, but on innoc ent p arties;
it re quires that aggregate c o llateral damage not be “too great” relativ e to the def ensiv e gain. Se e M cM ahan,
“Humanitarian Interv ention,” 22–24.
21
A mix e d/ c ombine d justific ation f or def ensiv e harm is one in which some of the harm enjo ys a liability
justific ation and some of the harm enjo ys a lesser-evil justific ation. (E.g., it ma y be permissib le to break both
legs of an attacker who is only liab le to suff er one bro ken leg, in virtue of the f act that there is a lesser-evil
justific ation f or breaking the e xtra leg.) Se e Jeff M cM ahan, “What Rights M a y be Def ende d b y M eans of W ar?”
The Mor a lity of Defensive W a r , e ds. Cé cile F abre a nd Seth Lazar (Oxf ord: Oxf ord Univ ersity Press, 2014):
133–34.
69
prote ction.
Ac c ording to Exclusion , T arget’s refusal of prote ction makes it so that the f act that
pu nc hi ng Attac ker wou ld save T a rget’s life is no longer a reason f or Def ender to punch At-
tacker. This makes it so that there is no longer sufficient reason to justify punching Attacker,
thus undermining Def ender’s permission to do so . As w e note d previously (§2.3.3), this is
an i mp lausib le imp lic ation.
The good news f or Contraction is that it does not ha v e this imp lic ation. This is be c ause
Contraction does not e x clude T arget’s interests from Def ender’s poo l of reasons. On this
ac c ount, T arget’s refusal of prote ction makes it so that Attacker is not liab le to be harme d b y
Def ender—which is to sa y that Def ender w ould infringe Attacker’s rights b y lightly punch-
ing him. But in a c ase where Def ender c an prev ent T arget from being kille d with a light
punch, the reasons to punch (i. e., sa ving T arget’s lif e) are so great as to outw eigh Attacker’s
right against this punch. Punching Attacker w ould be a permissib le rights-infringement.
22
Similar c onsiderations app ly in group c ases. Consider, f or e x amp le:
Grenade Defense : Attacker launches an attack. If suc c essful, each of one hun-
dre d pe op le will be kille d. Only Def ender c an stop Attacker, and he c an do so
only b y thro wing a grenade at him. The b last will kill Attacker, but it will also
22
But c onsider the f o llo wing v ariation on Easy Defense : Def ender c an prote ct T arget either b y punching At-
tacker or b y punching an innoc ent b ystander; punching the Attacker will set b ack his interests slightly more
than punching Bystander will set b ack hers. Does Contraction imp ly that Def ender should punch By-
stander rather than Attacker? It does not. Contraction tells us that Attacker is like Bystander in not being
liab le to be punche d. But this doesn’t mean that Def ender has no more reason to punch Attacker than By-
stander. Considerations of distributiv e f airness are an important sourc e of reasons, and responsibility makes a
diff erenc e to f airness. Be c ause Attacker is responsib le f or the f act that Def ender must choose betw e en diff erent
distributions of harm (let T arget die; punch Attacker; punch Bystander), it is more f air that Attacker suff er a
punch than Bystander. This giv es Def ender a strong reason to punch Attacker rather than Bystander ev en if
neither person is liab le to be so punche d.
70
kill an innoc ent b ystander. Ninety-nine members of the target group refuse
prote ction.
On Exclusion w e don’t c ount the interests of those who refuse prote ction as a reason
to thro w the grenade. Among the target group , only the interests of the lone non-refuser
c ontribute to the poo l of reasons f or thro wing the grenade. But ev en if w e entirely disc ount
the interests of Attacker,there is a v ery strong reason to not thro w the grenade: it will kill an
innoc ent b ystander. Ac c ording to Exclusion , then, Def ender thus f ac es a choic e betw e en
a l lowi ng one innoc ent person to be kille d and ki l li ng an innoc ent person. It is not normally
permissib le to do some harm in order to a v ert an e quiv alent (or lesser) harm. On Exclusion ,
then, Def ender ma y not thro w the grenade.
Contraction , ho w ev er, deliv ers the more p lausib le v erdict: it permits Def ender to thro w
the grenade. The interests of the ninety-nine refusers are not e x clude d. And that’s a lot of
liv es! The benefits of sa ving one-hundre d liv es is more than enough to giv e us a T ro lley-style,
lesser-evil justific ation f or killing one person as a side-eff e ct. Contraction thus lea v es us
with a c ombine d justific ation f or thro wing the grenade: there is a liability justific ation f or
the harm to the Attacker, and a lesser-evil justific ation f or the harm to the b ystander. Again,
this result strikes me not just as an interesting imp lic ation, but as a virtue of the ac c ount. It
is v ery intuitiv e that Def ender is permitte d to thro w the grenade.
2.5.4 Military Intervention
What does all of this mean f or the ethics of military , humanitarian interv entions? The orists
like Helen Fro w e
23
and Jeff M cM ahan
24
ha v e claime d that the refusal of prote ction is all but
23
Fro w e, “Judging Arme d Humanitarian,” 10 8–09.
24
M cM ahan, “Humanitarian Interv ention, C onsent,” 51–53.
71
irrelev ant to jus ad be l lu m c onsiderations, pre cisely be c ause (i) the refusal of prote ction is
nev er unanimous in real-w orld humanitarian c rises and (ii) less-than-unanimous refusal does
not(on its o wn) undermine the permissibility of interv ention.
P arry’s Exclusion ac c ount is much more bullish on the po w er of refusal. It reje cts (ii),
claiming that the permissibility of prote ctiv e, military interv ention c an be undermine d b y
less-than-unanimous refusal. In f act, refusal b y ev en a small minority of the target group c an
make the diff erenc e. Imagine that nation D must choose whether to def end nation T from
nation A ’s unjust aggression. Interv ention will c ost the liv es of man y so ldiers from nations D
and A, as w ell as the liv es of man y non-c omb atants from nation T . Giv en these c osts, suppose,
interv ention w ould just ba re ly clear the b ar of justific ation in the absenc e of an y refusal from
the pe op le of nation T ; the b alanc e of reasons tilts ev er so slightly in f a v or of interv ention.
In these (not unrealistic) c onditions, only a small number of the citizens of nation T ne e d
refuse prote ction to undermine the justific ation f or interv ention. By refusing prote ction,
they w ould re duc e the reasons in f a v or of interv ention, tipping the justific atory b alanc e.
Contraction is not nearly so bullish. It ho lds that less-than-unanimous refusal ca n ren-
der interv ention impermissib le. But realistic c ases where this happens will be much harder to
c ome b y f or Contraction than f or Exclusion , pre cisely be c ause Contraction ho lds
that refusal on ly makes a diff erenc e to c onsiderations of na rrow proportionality . Return to
the c ase in which nation D’s interv ention w ould just b arely clear the b ar of jus tific ation in
the absenc e of an y refusal from the pe op le of nation T . Imagine that a small number of the
citizens of T refuse prote ction. Ac c ording to Contraction , these citizens do not remo v e
their interests from the e quation; they do not make it harder to satisfy c onsiderations of wide
proportionality . What they do , rather, is make it so that their rights no longer c ontribute to
72
the liability of the c omb atants of nation A. But ev en if w e ignore the c ontribution to liability
made b y t hose rights, the individual c omb atants of nation A will in most c ases remain liab le
to v ery serious harm simp ly in virtue of their serious threats to the rights of the large, non -
refusing subset of the population. This means that the refusal of prote ction b y the minority
will not usually entail that some one is wronged b y interv ention who w ouldn’t otherwise be
wronge d.
So while Contraction does not quite treat refu sal as altogether irrelev ant in the c onte xt
of w ar, it does not grant it nearly the po w er that Exclusion does. E v en when interv ention
w ould b arely clear the justific atory b ar in the absenc e of refusal, Contraction tells us that
refusal b y a minority of the target nation will not make the diff erenc e when the liability of the
c omb atants of the aggressor nation is hea vily o v erdetermine d. But in the real w orld, where
military aggression threatens the rights of very ma n y citizens, c omb atant liability is almost
alw a ys hea vily o v erdetermine d.
2.6 State Officials
Bef ore c oncluding, I w ant to briefly c onsider a challenge c onc erning our po w er to refuse
prote ction from our o wn state officials. In practic e, prote ction from aggression often c omes
from such agents — from, f or e x amp le,the po lic e. Consider a v ariant of Si b li ngs :
Si b li ngs 2 . As bef ore, e x c ept that an on-duty po lic e offic er, P am, is also present,
and c an def end T eresa b y breaking Adam’s leg. T eresa refuses both Dan and
P am’s prote ction.
I think man y readers will share m y intuition that P am’s normativ e situation is diff erent
73
from Dan’s. Whereas it se ems that Dan should not interv ene against T eresa ’s wishes, it se ems
P am should. Ho w to e xp lain this diff erenc e?
Let me off er an e xp lanation f or this diff erenc e that most an y ac c ount of refusal c an help
itself to . (Cases like Si b li ngs 2 pose a challenge f or an ac c ount like Exclusion just as much
f or Exclusion .) In a democ ratic po litic al c ommunity , the members of the c ommunity o w e
it, not just to the state, but to one a not her to obey those la ws (at least in so f ar as those la ws
are just). This means that when some one breaks the la w , they do not just wrong those who
are harme d b y that la w-breaking; they wrong ev ery member of the c ommunity .
No w these rights against la w-breaking ha v e an important diff erenc e from our more fun-
damental rights against harm. When A threatens T’s rights against harm, A makes himself
univ ersally liab le. That is, he makes himself liab le to be harme d b y an y potential def ender.
Not so , I think, in the c ase of our rights against la w-breaking. In modern po litic al c ommuni-
ties, w e ha v e chosen c ertain state officials — our “la w enf orc ement” officials — to be the ones
to enf orc e our rights against la w-breaking. What this means, I claim, is that these officials ha v e
both a duty to enf orc e the la w and a spe cial permission to do so . Some one who attempts to
break the la w is not liab le to be harme d b y just an y one to prev ent his la w-breaking; he is liab le
( qua la w-breaker) only to def ensiv e harm at the hands of la w enf orc ement officials.
25
This, I claim, largely e xp lains wh y refusal does not usually undermine the permissibility of
po lic e interv ention, ev en when it does undermine the permissibility of civilian interv ention.
When T eresa refuses prote ction, she mitigates the liability eff e cts of the threat to her rig ht
ag a i nst ha rm . But b y breaking Adam’s leg, P am does not just prev ent a threat to t his right.
25
Of c ourse, man y c ases in which A poses a threat of breaking the la w are also c ases in which A thereb y poses
a threat to some one’s rights against harm. If (ne c essary and proportionate) def ensiv e harm c an prev ent both
f orms of wrongdoing in such a c ase, A is theref ore univ ersally liab le to such harm. M y claim is just that the mere
threat of la w-breaking grounds only a liability to prev entativ e harm at the hands of la w enf orc ement officials.
74
Si b li ngs 2 is a spe cial c ase of group refusal (§2.5.2) in which only one member of a target
group refuses prote ction. F or in addition to his threat to T eresa, Adam also poses a threat
to the rights of ev ery member of the po litic al c ommunity that he not break the ir c riminal
la ws against b attery . In virtue of t his threat, Adam is liab le to c onsiderab le harm from P a m
to prev ent the b attery . But he is not likewise liab le relativ e to Da n , f or the reason that the
threat of mere la w-breaking only renders a person liab le to be harme d b y those our po litic al
c ommunity has designate d to enf orc e the la w .
The rights each of us ha v e against one another that w e obey our la ws — together with the
sele ctiv e permissions our po litic al institutions c onf er on la w enf orc ement officials to prev ent
the infringement of these rights — is what e xp lains wh y refusal ’s magic is less potent when
dire cte d to w ard those officials.
2.7 Conclusion
One of the chief po w ers of refusal c onc erns our po w er to mitigate the normativ e eff e cts of
our o wn rights — spe cific ally , our po w er to c ontract the set of persons with respe ct to whom
one’s attacker is liab le.
26
W e e x ercise this po w er when w e refuse prote ction from an attacker.
Refusal remo v es the attacker’s liability with respe ct to the person whose prote ction is re-
26
This is not the only thing that refusal does, but it is the most important f or understanding ho w refusal
c an render prote ction impermissib le. As I indic ate d to w ards the beginning of §2.4, the refusal of prote ction
surely also c ontributes other reasons ag a i nst interv ention, in virtue of the f act that an act of refusal e xpresses
the refuser’s pref erenc es, v alues, and c ommitments. But refusal will sometimes also remo v e c ertain reasons
for interv ention. M ost clearly: when X refuses Y’s prote ction this will typic ally release Y from an y duty of
prote ction he ma y ha v e o w e d X. (Thanks to Michelle Dempsey f or helpful discussion on this point.)
75
fuse d.
27
And be c ause an attacker’s liability is typic ally a ne c essary ingre dient f or the w ould-be
def ender’s permission to harm him, refusal thus often suc c e e ds in undermining that permis-
sion.
The Liability Contraction Ac c ount, as I’v e c alle d it, does not just giv e us a diff erent
b ackstory f or the moral eff e cts of refusal. It also giv es us a v ery diff erent ac c ount of what those
eff e cts are. W e’v e c omp are d it against its only dev elope d riv al, the Exclusion Ac c ount. In
some respe cts Contraction attributes less po w er to refusal than Exclusion , and in some
respe cts more. On Contraction , refusal is less eff e ctiv e at rendering prote ction impermis-
sib le, sinc e refusal does not prev ent the refuser’s interests from c ontributing to a lesser-evil
(or “mix e d ”) justific ation f or prote ction (§2.5.3). But Contraction also attributes to re-
fusal an eff e ct that Exclusion does not: it grants to some minimally-responsib le attackers
the permission to def end themselv es from impermissib le, prote ctiv e interv entions (§2.5.1).
27
The po w er w e e x ercise b y refusing prote ction ma y be ev en more fine-graine d than this. Sometimes our
refusal is mea ns-specifi c . W e sa y , “Y ou c an prote ct me in w a y X, but not w a y Y .” When w e issue such means-
spe cific refusal, I claim, w e restore only some of the attacker’s rights with respe ct to the person whose prote ction
w e refuse. W e restore the attacker’s right not to suff er def ensiv e harm in w a y Y without restoring his right not
to suff er def ensiv e harm in w a y X.
76
3
Chapter Thre e: Def ense W ithout Threat
3.1 Introduction
Suppose y ou ne e d a liv er transp lant. Y ou’ll die without one. But the w aiting list is too long,
and the only w a y y ou c an get a liv er is if y ou kill y our ne xt-door neighbor and steal theirs.
That w ould be wrong!
Or suppose y ou are attacke d b y a gunman. The only w a y to prote ct y our lif e is to grab an
77
innoc ent b ystander and use them as a human shield, resulting in their death. This too w ould
p lainly be wrong.
As these c ases illustrate, it is normally wrong to intentionally harm others, ev en if b y harm-
ing them y ou c an prev ent c omp arab le harm to y ourself. But there are e x c eptions to this
rule. And one of the most w ell-re c ognize d e x c eptions is the c ase of self-def ense. Consider
a p aradigm e x amp le:
P a r adigm . A villain attempts to murder y ou. He will suc c e e d unless y ou kill
him first.
Y ou ma y not steal y our neighbor’s liv er or use a b ystander as a human shield to sa v e y our
lif e. But y ou ma y kill villain to sa v e y our lif e.
The diff erenc e is that only V illain has made himself lia b le to be harme d — that is, he has
done something to lose some of his normal rights not to be harme d.
1
Y our neighbor and the
b ystander ha v e done no such thing. They retain their normal rights not to be harme d. This
is wh y y ou are not permitte d to harm them. But no such rights stand in the w a y of harming
V illain — henc e y our permission to harm him.
Liability is t he c entral ingre dient in most justific ations f or the use of def ensiv e f orc e.
2
And
so a the ory of liability must be at the c enter of our the ory of the ethics of def ense. If w e w ant
1
In the present p aper, I’ll use the terms ‘liab le’ and ‘liability’ in a more e xp ansiv e w a y than they are sometimes
use d. Some the orists use the term in such a w a y that one is not “liab le” to some harm when they’v e lost their
right against that harm b y w a y of an act of c onsent (se e, f or e x amp le, T adros, “Duty and Liability ,” Uti litas 24
(2012): 260). As I’m stipulativ ely using the term, ho w ev er, it doesn’t matter ho w one loses the right in question.
On m y usage, f or a person to be liab le to φ just is f or them to lack their normal right against φ .
2
I sa y ‘most’ be c ause some harm impositions are permissib le rights-infringements, as when y ou re dire ct
the tro lley a w a y from killing fiv e pe op le and onto a p ath where it will break one person’s f oot. Such harm
impositions admit of a so-c alle d “lesser-evil ” justific ation rather than a liability justific ation. M an y the orists also
endorse so-c alle d “agent-relativ e-prerogativ e” justific ations. Se e, f or e x amp le, Jonathan Quong, The Mor a lity
of Defensive Forc e (Oxf ord: Oxf ord Univ ersity Press 2020): 58–96.
78
a t he ory of def ense, w e ne e d a the ory of when and wh y some one loses rights against harm.
So what is it that makes some one like V illain liab le to be harme d? W ell it sure se ems to
ha v e something to do with the f act that he wou ld ha rm someone if he isn’t ha rmed hi m fi rst .
That se ems, at first glanc e, at least a ne c essary c ondition f or liability .
But there’s a serious and long-standing prob lem with this idea, which is that there are c om-
pelling c ases where a person se ems liab le to harm ev en if they pose no actual threat.
3
Suppose
some one attempts to kill y ou, but unbekno wnst to either of y ou their gun is jamme d.
4
Or
suppose some one c on vincingly pretends to attempt to kill y ou, so as to giv e y ou the sc are of
y our lif e.
5
Neither the “futile attempter” nor the “b luff er” pose an y genuine threat of harm.
And y et it sure se ems they are liab le to be harme d.
Cases like these pose a serious challenge f or the the ory of liability . W e ne e d a the ory that c an
ac c ount f or our judgments in such c ases — but that does not at the same time over-gener ate
c ases of liability . After all, one is not rendere d liab le merely b y a ppea ri ng to pose a wrongful
threat. Suppose, f or e x amp le, that Bloggs re c eiv es c ompelling, but f alse, testimonial evidenc e
that y ou’re attempting to murder him. Y ou thus appear to him to pose a wrongful threat.
Bloggs ma y ev en ha v e an evidenc e-relativ e justific ation f or harming y ou. But y ou p lainly are
not liab le to def ensiv e harm.
I’ll c all this challenge the Cha l lenge of Mere ly A ppa rent Attac kers . I think it’s a challenge
of c entral importanc e to the the ory of liability , not least be c ause the relev ant f eatures of these
c ases are f ar from e x otic. On the c ontrary: in the real-w orld, def ensiv e agents v ery often oper-
3
Se e, f or e x amp le, Alberic o Gentili, De Ju re Be l li (On t he Law of W a r) , trans. Jo hn C. Ro lf e (Clarendon
Press 1933)(1598): Bk. I, ch. 14, p . 62–63.
4
This c ase is inspire d b y Helen Fro w e’s “ App arent Murderer” c ase from Defensive Ki l li ng (Oxf ord 2014):
85.
5
This c ase is inspire d b y Kimberly Kessler F erzan’s c ase b y the same name from “The Bluff: The P o w er of
Insinc ere Actions,” Leg a l Theory 23 (2 017):169.
79
ate under c onditions of f alse inf ormation. Real-w orld def ensiv e agents fre quently ha v e mis-
taken beliefs about all sorts of things: about the threat others pose, about what def ensiv e
options they ha v e a v ailab le, and about the potential c onse quenc es of those options. As such,
an answ er to the Challenge is an important p art of an ethics of def ense that speaks to the c on-
duct of def ensiv e agents in the real w orld — be they so ldiers, po lic e, or priv ate individuals.
As I’ll sho w in this chapter, it’s also an important challenge be c ause of what w e learn about
the ethics of def ense in answ ering it. T o date, the most f ocuse d attempts to answ er this chal-
lenge all proc e e d b y positing an additional ground of liability: there’s one ground of liability
f or attackers who pose an actual threat of harm and another ground of liability f or merely
app arent attackers.
6
I agre e that the challenge sho ws the ne e d f or a more nuanc e d the ory of
liability . But I disagre e that the nuanc e c onc erns the grou nds of liability . In what f o llo ws,
I def end an answ er to the Challenge of M erely App arent Attackers where the nuanc e c on-
c erns, not the grounds, but the properti es of liability . I’ll argue f or the surprising claim that
merely app arent attackers are not , in f act, strictly liab le to def ensiv e harm, but that w e c an
nonetheless e xp lain wh y some, and only some, merely app arent attackers bear the “upshots ”
of liability .
§3.2 more c arefully presents the terms of the challenge. §3.3 summarizes thre e e xisting
answ ers to the challenge, setting the stage f or m y answ er in §3.4. §3.5 unp acks important
adv antages of m y ac c ount. §3.6 c oncludes.
6
F erzan, “The Bluff”; Fro w e, Defensive Ki l li ng : 85-86; René e Jorgensen, “The M oral Grounds of Rea-
sonab ly Mistaken Self-Def ense,” Phi losop hy a nd Phenomeno logi ca l Resea rc h 103 (2021):140-156; M cM ahan,
“Who is M orally Liab le to be Kille d in W ar?” A na lysis 71 (2011): 555-556.
80
3.2 The Challenge of Merely Apparent At tackers
Consider tw o c ases, one in v o lving a “futile attempter” and the other a “b luff er”:
Ja m . Jamm y intends to murder Def ender. He points a gun at her and prep ares
to pull the trigger. Def ender kno ws she c an prev ent Jamm y from pulling the
trigger only b y breaking his leg. Believing that she will be kille d otherwise, De-
f ender breaks Jamm y’s leg. Unbekno wnst to an y one, ho w ev er, Jamm y’s gun is
irrep arab ly j amme d.
Bluff . Bluff er de cides to p la y a v ery ill-c onc eiv e d prank on his w orkp lac e man-
ager, Def ender. He brings to the offic e an unlo ade d gun. During a me eting he
stands up , points the gun at Def ender, and y ells, ”E at lead!”. Def ender reason-
ab ly believ es the gun is real and that she is about to be kille d. Kno wing she c an
prev ent Jamm y fr om pulling the trigger only b y breaking his leg, she does so .
Jamm y and Bluff er pose no actual threat to Def ender. Def ender w ould not be harme d w ere
she to abstain from harming Jamm y or Bluff er. And y et: Jamm y and Bluff er sure se em liab le
to ha v e their leg bro ken b y Def ender. At least they bear the usual u ps hots of liability . Let me
e xp lain. When some one has a right not to be harme d, they are typic ally permitte d to def end
against threats to that right b y f orc e, others are typic ally permitte d to def end against threats
to that right b y f orc e, and c ompensation is typic ally o w e d if the right is in f act vio late d.
7
W e
se e this clearly in the P aradigm c ase. Y our rights are threat ene d and y ou c an fight b ack against
7
At least, def ense against a threat to one’s rights is typic ally permissib le so long as it does not vio late the
ne c essity or proportionality c onstraints.
81
V illain, others c an fight b ack on y our behalf, and y ou w ould be o w e d c ompensation w ere
V illain to suc c e e d in injuring y ou.
But no w c onsider the position of V illain — the position of some one who has made himself
liab le to harm. Suppose y ou fight b ack against him. Is he permitte d to fight b ack in turn —
to engage in c ounter-def ense? Certainly not. Likewise, a b ystander is not permitte d to fight
b ack against y ou on his behalf. And if y ou do injure him in self-def ense, y ou don’t o w e him
an y c ompensation.
So these are thre e typic al upshots of liability . If y ou are liab le to ha v e some harm impose d
on y ou, then it will typic ally be true that:
(U1): Y ou w ould not be permitte d to fight b ack against that harm (at least not with c omp a-
rab le or gre ater f orc e),
8
(U2): Others w ould not be permitte d to fight b ack on y our behalf (again, at least not with
c omp arab le or g reater f orc e), and
(U3): Y ou w ould not be o w e d c ompensation if y ou w ere to suff er that harm.
Jamm y and Bluff er bear all thre e of these upshots. They w ouldn’t ha v e be en permitte d to
fight b ack against Def ender (they w ouldn’t ha v e be en permitte d, f or e x amp le, to break De-
f ender’s leg to prev ent her from breaking theirs.) Likewise, third-p arties w ould not ha v e be en
permitte d to fight b ack on their behalf. And Jamm y and Bluff er aren’t o w e d c ompensation
f or their bro ken legs.
8
I insert the p arenthetic al qualifier in order to lea v e open the possibility that a person who is liab le to be
targete d with sev ere harm H ma y , under some circumstanc es, be permitte d to def end himself with on lesser-evil
grounds with v ery mild c ounter-def ensiv e harm.
82
The lesson of the Jam and Bluff c ases is that a person c an bear the upshots of liability ev en
if they pose no actual threat — ev en if they are what I’ll c all a mere ly a ppa rent attac ker . W e
ne e d a the ory of liability that e xp lains wh y .
But w e also ne e d a the ory that c an properly disc riminate betw e en c ases where merely ap-
p arent attackers bear the upshots of liability and c ases where they do not . Jam and Bluff are
tw o c ases where merely app arent attackers bear the upshots of liability . Here are tw o c ases
where the merely app arent attacker does not:
F a lse T esti mon y . T estifier w ants to se e Innoc ent harme d b y Def ender, and so
he lies to Def ender: he tells her that Innoc ent is about to kill her. Def ender
has e x c ellent reason to trust T estifier’s testimon y , and so she attacks Innoc ent,
breaking hi s leg.
9
Evi l Twi n. While on a ro ad trip , T win’s engine o v erheats. He w alks to the near-
est to wn and enters the first me chanic shop he c omes ac ross. Unbekno wnst
to T win, ho w ev er, he has an evil twin brother — a notorious serial killer who
has just esc ape d from prison. Authorities ha v e w arne d loc als that the killer will
shoot an y one he c omes ac ross on sight. Believing T win to be the murderer and
believing himself to be in imminent danger, the me chanic, Def ender, lunges at
T win with a c ro wb ar, breaking his leg.
10
As in Jam and Bluff, the def ender in F alse T estimon y and E vil T win attacks some one who
merely appears to pose a wrongful threat. But where Jamm y and Bluff er bear the upshots
9
This c ase is inspire d b y Michael Otsuka ’s “Dignitary” c ase from “Killing the Innoc ent in Self-Def ense,”
Phi losop hy a nd Pu b li c Aff a i rs 23 (1994):91.
10
This c ase is inspire d b y Jeff M cM ahan’s “Mistaken Resident” c ase from “Basis of M oral Liability”: 387,
and Jonathan Quong’s “Mistaken Attacker” c ase from Defensive Forc e : 23.
83
of liability , T estifier and T win do not. They w ould ha v e be en permitte d to fight b ack against
Def ender, third-p arties w ould ha v e be en permitte d to fight b ack on their behalf, and they a re
o w e d c ompensation f or their bro ken legs.
11
So here’s the Cha l lenge of Mere ly A ppa rent Attac kers . W e ne e d a the ory of liability that
c an c apture the intuitiv e diff erenc es betw e en characters like Jamm y and Bluff er (on the one
hand) and characters like Innoc ent and T win (on the other). W e ne e d a the ory that e xp lains
wh y the f ormer bear the upshots of liability , but not the latter.
3.3 Culpability, Responsibility, and Signaling
There ha v e be en a handful of attempts to answ er the Challenge of M erely App arent Attack-
ers, all of which proc e e d b y adding an addendum to an ac c ount of liability f or actual attackers
that has the eff e ct of attributing liability to characters like Jamm y and Bluff er (and not Inno-
c ent and T win).
One of the most most f ocuse d and dev elope d attempts c omes c ourtes y of Kimberly Kessler
F erzan. On her view , what makes an actua l attacker liab le to be harme d is that they are cu l-
pa b le f or the f act that they w ould otherwise vio late some one’s right not to be harme d.
12
Cul-
p ability also p la ys a starring ro le in e xp laining wh y merely app arent attackers c an be liab le
to harm. What makes them liab le, of c ourse, isn’t that they are culp ab le f or posing an ac-
11
This judgment appears to be widely , though not univ ersally , share d. F or e x amp les of at least p artial
endorsement, se e Sab a Bazargan-F orw ard, “Def ensiv e Liability W ithout Culp ability ,” in Christian Coons &
Michael W eber (e ds.), The Et hi cs of Se lf- Defense (Oxf ord Univ ersity Press 2016): 72; M cM ahan, “The Basis of
M o ral Liability”: 387; Otsuka, “Killing the Innoc ent”: 91; Quong, Defensive Forc e : 24. In c onsidering a c ase
like E vil T win, Larry Al e x ander appears to take the view that T win w ould be permitte d to fight b ack, but that a
b ystander w ould not be permitte d to fight b ack on T win’s behalf. Se e his “Re cipe f or a The ory of Self-Def ense,”
in C hristian Coons & Michael W eber (e ds.), The Et hi cs of Se lf- Defense (Oxf ord Univ ersity P ress 2016):29.
12
“W ith respe ct to the p aradigmatic instanc e of self-def ense, an aggressor be c omes liab le to the def ender’s
f orc e be c ause he has culp ab ly de cide d to harm the def ender and the def ender must use f orc e to stop the harm
from oc curring” (F erzan, “The Bluff,” p . 173).
84
tua l threat of harm (b y definition they pose no such threat). Rather, they are liab le be c ause
they are culp ab le f or a ppea ri ng to pose a threat (i. e., f or the def ender’s be li ef that they pose a
threat).
13
W e’ll c all F erzan’s the Culpability Account of liability .
Culp ability : What makes a person liab le to (ne c essary and proportionate)
def ensiv e harm (or app arent def ensiv e harm) is that either (i) they are culp ab le
f or the f act that some one’s rights w ould otherwise be vio late d, or (ii) they are
culp ab le f or the f act that the def ender (reasonab ly) believ es some one’s rights
w ould ot herwise be vio late d.
On this ac c ount, Jamm y and Bluff er bear the upshots of liability be c ause they are in f act
liab le. And they are liab le be c ause they are culp ab le f or appearing to pose an actual, wrongful
threat. Innoc ent and T win, b y c ontrast, bear no culp ability f or their appearing to pose a
threat.
Jeff M cM ahan and (sep arately) Helen Fro w e off er an ac c ount that is structurally analogous
to F erzan’s, but that rep lac es the notion of cu lpa bi lity with that of responsi bi lity . Call this the
Responsibility Account :
Responsibility : What makes a person liab le to (ne c essary and proportion-
ate) def ensiv e harm (or app arent def ensiv e harm) is that either (i) they are re-
sponsib le f or the f act that some one’s rights w ould otherwise be vio late d, or (ii)
they are responsib le f or the f act that the def ender (reasonab ly) believ es some-
13
Ibid, p . 172.
85
one’s rig hts w ould otherwise be vio late d.
14
As on F erzan’s ac c ount, the M cM ahan-Fro w e ac c ount tells us that Jamm y and Bluff er bear
the upshots of liability be c ause they are in f act liab le. And they are liab le be c ause they are
responsib le f or appearing to pose an actual, wrongful threat. Innoc ent and T win, b y c ontrast,
bear no responsibility f or their appearing to pose a threat.
Another attempt to answ er the Challenge c omes c ourtes y of René e Jorgensen. Jorgensen
does not purport to off er a c omp lete the ory of liability . Rather, she off ers an addendum that
is meant to be c omp atib le with man y diff erent the ories of liability in actual-threat c ases. So
f or Jorgensen, there’s the usual c onditions f or liability that w e find satisfie d in actual-threat
c ases — whatev er e x actly those might be. But then there’s a se c ond p ath to liability . Instead
of me eting the usual c onditions f or liability y ou might make y ourself liab le b y engaging in
beha vior that c on v entionally signa ls that y ou me et the usual c onditions f or liability . (T o il-
lustrate: suppose w e attache d this addendum to Thomson’s ac c ount of liability , ac c ording to
which an actual attacker is liab le when and be c ause he w ould vio late some one else’s rights if
he w ere not harme d himself. The result w ould be that a merely app arent attacker makes him-
self liab le when and be c ause he c on v entionally signals that he w ould vio late some one else’s
rights if he w ere not harme d himself.) I’ll c all Jorgensen’s the Signaling Account :
Sign aling (Jorgensen): What makes a person liab le to (app arent) def ensiv e
harm is that either (i) they me et the usual c onditions, C, f or liability (i. e., the
14
M cM ahan, “Who is M orally Liab le?”: 555-556; Fro w e, Defensive Ki l li ng : 86-86. The abo v e f ormulation
glosses o v er the f act that, at least f or M cM ahan, it is c ompa r ative responsibility that matters most. This is the
c rucial diff erenc e betw e en Jamm y /Bluff er and Innoc ent/T win: Jamm y /Bluff er bear more responsibility than
Def ender f or her belief that they pose a threat, whereas Def ender bears more responsibility than Innoc ent/T win
f or her mistaken belief about them.
86
c onditions spe cifie d b y the c orre ct the ory of liability in p aradigmatic, actual-
threat c ases), or (ii) they ha v e perf orme d an action that c on v entionally signa ls
to othe rs that p , where c onditions C w ould be satisfie d if p w ere true.
Jorgensen thus also ho lds that Jamm y and Bluff er bear the upshots of liability be c ause
they are in f act liab le. They are liab le be c ause they engage in beha vior (e.g., pointing a gun at
some one, y elling ”E at lead!”) that c on v entionally signals that they pose an actual threat.
W e’ll c ome b ack to these the ories later in the p aper. F or no w I mention them as b ack-
ground against which to c ontrast m y o wn strategy f or answ ering the Challenge. The strategy
emp lo y e d b y the abo v e the orists is to posit a se c ond p ath to liability , alongside the p ath tra v-
ele d b y actual attackers. Jamm y and Bluff er bear the upshots of liability in virtue of tra v eling
this se c ond p ath. They are not liab le f or the same reasons that actual attackers are liab le, but
they are liab le nonetheless.
I think there is a great deal to learn from these ac c ounts about the importanc e of respon-
sibility , culp ability , and signaling beha viors. And inde e d, as w e’ll se e, there is much m y o wn
answ er to the Challenge dra ws from these ac c ounts. Nonetheless, I think the abo v e answ ers
to the Challenge are mistaken in tw o respe cts. First, they are each inc omp lete: they only tell
p art of a more c omp le x story . Se c ond, they f ail to c apture c entral diff erenc es betw e en the
moral position of actual attackers as c omp are d to merely app arent attackers.
M y o wn answ er to the Challenge takes a v ery diff erent tack. Rather than proposing a
se c ond p ath to liability , I’ll present an ac c ount of liability ac c ording to which there is only
one p ath. On this ac c ount, Jamm y and Bluff er are not liab le to harm, strictly speaking. But
the ac c ount nonetheless e xp lains wh y they bear the thre e upshots of liability . In §3.4 I’ll
unp ack this ac c ount of liability and the answ er it off ers to the Challenge. In §3.5 I’ll e xp lain
87
wh y I pref er this answ er.
3.4 A New Answer
3.4.1 The Assumption Account
Let me introduc e m y ac c ount b y first sa ying something briefly about the structure of liabil-
ity . A person is nev er liab le to harm si mp li citer . Liability is alw a ys person-relativ e and go al-
relativ e.
15
When a person is liab le to be harme d, they are liab le to be harme d b y so-a nd-so as
a mea ns or si de-effect of su c h-a nd-su c h goa l . That’s the b asic structure of the liability relation.
Briefly , here’s wh y w e should think t hat liability is person-relativ e. Liability , on m y usage,
is just the absenc e of a right. It is clearly possib le to ha v e a right that one person not harm y ou
in a c ertain w a y but to lack that same right against some one else. W e se e this clearly in c ases
in v o lving c onsent. If I c onsent to a bo xing match with Alic e but not Bob , then I am liab le to
c ertain harms at the hands of Alic e but not Bob .
M ore important f or present purposes, ho w ev er, is the idea that liability is go al-relativ e.
Wh y think that’s true? W ell c omp are:
Rec kless Drivi ng 1 . Driv er is re cklessly driving abo v e the spe e d limit in an eff ort
to make his dinner reserv ations. He loses c ontro l of his c ar. It hurtles to w ards
Def ender. Def ender will be kille d b y the c ar unless she kills Driv er b y re dire cting
his c ar a w a y fr om her and into a tre e. (Driv er will be unharme d otherwise.)
Rec kless Drivi ng 2 . As bef ore, Driv er’s c ar hurtles to w ards Def ender. This time,
15
Here I agre e with (among others) Jeff M cM ahan. M cM ahan claims that liability is go al-relativ e in Ki l li ng
i n W a r , (Oxf ord 2009): 8-9, and that liability is person-relativ e in “The Limits of Self-Def ense,” in Christian
Coons & Michael W eber (e ds.), The Et hi cs of Se lf-Defense , (Oxf ord 2016): 201-203.
88
ho w ev er, Def ender c an simp ly step out of the w a y , in which c ase Driv er will be
unharme d. But Def ender also has the option to re dire ct Driv er’s c ar, in which
c ase his body will shield a b ystander from the gunfire of a villain in the neigh-
borhood (t his will kill Driv er).
In either c ase, killing Driv er is a def ensiv e harm; in either c ase, killing Driv er sa v es the lif e
of some one who is in danger. And y et it is permissib le f or Def ender to kill Driv er only in the
first c ase.
The best e xp lanation f or this diff erenc e is that Driv er’s liability is spe cific ally tie d to the
go al of a v erting his own t hreat . He is liab le to be kille d in servic e of t hat go al, but not in
servic e of the go al of a v erting the villain’s attack.
16
I think the best w a y to make sense of the go al-relativity of liability is b y appeal to an idea
introduc e d b y V ictor T adros.
17
It’s the idea that a person c an lose c ertain claim rights b y
taking on c ertain kinds of duties. Duties ha v e the right structure to do the w ork of e xp laining
the go al-relativity f eature, sinc e duties are themselv es go al-relativ e. Duties are alw a ys duties
to do somet hi ng , to see to some goa l .
The idea, more pre cisely:
Liability b y Duty . When X has not c onsente d to be harme d but is nonetheless
liab le to be harme d, this is be c ause (i) X has a duty to se e to go al G and (ii)
harming X is a me ans or side-eff e ct of bringing about G.
16
I’m not claiming that Driv er c an nev er be liab le to harm that happens to a v ert the villain’s attack. F or
e x amp le, if Def ender’s only w a y to prote ct herse lf in v o lv e d re dire cting Driv er such that he w ere to shield the
b ystander, then Driv er wou ld be liab le to be so harme d. But he w ouldn’t bear this liability in virtue of the f act
that the harm is ne c essary to prote ct b ystander. He w ould bear this liability in virtue of the f act that the harm
is ne c essary to prote ct Def ender.
17
Se e his “Duty and Liability” and The Ends of Ha rm (Oxf ord 2011): 169-196. I also dev elop this idea in
m y “The Demands of Ne c essity ,” Et hi cs (f orthc oming).
89
Here’s ho w the idea p la ys out in actual-threat c ases like P aradigm. V illain has a b asic duty
not to harm y ou. Onc e he initaties his attack, this duty giv es rise to a se c ondary duty to se e
that t he c osts of his attac k not f a l l on a n yone but hi mse lf (sinc e otherwise he w ould f ail a duty
not to harm). This is wh y he is liab le to be harme d b y you : he is just being made to do what
he has a duty to do himself — namely , to se e that the c osts of his attack not f all on an y one
other than himself. His duty to assume the c osts of the attack makes him liab le to be made
to assume the c osts of the attack.
18
Call this the Assumption Account of l iability .
19
3.4.2 Assuming the Costs of Another’s Attack
No w in a c ase like P aradigm, V illain’s b asic duty not to harm is what e xp lains wh y he takes on
a duty to assume the c osts of the attack. But a duty not to harm is not the only thing that c an
ground a duty to assume the c osts of an attack. There are, in f act, a p lurality of w a ys in which
a person c an incur such a duty . Understanding this is the key to answ ering the Challenge of
M erely App arent Attackers.
Importantly , w e ne e d to notic e that there are w a ys a person c an take on a duty to assume
18
Here’s ho w the Liability b y Duty princip le helps to e xp lain the liability as ymmetry betw e en Re ckless Driv-
ing 1 and Re ckless Driving 2. Driv er’s duty not to ki l l Defender is much more demanding tha n his duty to
protect t he bysta nder from bei ng ki l led by t he vi l la i n . Driv er’s duty to w ards Def ender is so demanding that it
w ould re quire him ev en to re dire ct his c ar a w a y from her at the c ost of his o wn lif e (if this w ere the only w a y
to prev ent his killing her). Driv er’s duty to rescue the b ystander is not nearly so demanding. This diff erenc e in
the demandingness of Driv er’s o wn duties, I claim, is what e xp lains wh y he is liab le to be kille d in servic e of the
one go al but not the other.
19
This is close to T adros ’s o wn the ory of liability to def ensiv e harm. T adros writes: “One c entral w a y of
e xp laining a person’s liability to be harme d to a v ert a threat is to sho w that she has incurre d an enf orc eab le duty
to a v ert that threat, ev en if she will be harme d ” (“Causation, Culp ability , and Liability ,” in Christian Coons
& Michael W eber (e ds.), The Et hi cs of Se lf-Defense , (Oxf ord 2016):116. I agre e, but I’m claiming something
more general than this. At the heart of the present essa y is the idea that a person c an ha v e a duty to assume the
c osts of a vio lent enc ounter wit hout ha ving a duty to a v ert some wrongful threat. In f act, as w e’ll se e, a duty of
c ost assumption sometimes re quires the v ery opposite: it re quires a person to a bsta i n from a v erting a wrongful
threat.
90
the c osts of someone e lse ’s attack. W e se e this illustrate d in a v ariant of the F alse T estimon y
c ase:
Hu ma n Shi e ld . As bef ore, T estifier w ants to se e Innoc ent harme d b y Def ender,
and so he lies to Def ender: he tells her that Innoc ent is about to kill her. De-
f ender has e x c ellent reasons to trust T estifier’s testimon y , and so she attacks In-
noc ent ... As it turns out, ho w ev er, T estifier c an interv ene to def end Innoc ent,
but in only one of tw o w a ys: b y breaking Def ender’s leg or b y shielding Inno-
c ent such t hat T estifier’s o wn leg is bro ken.
Notic e, first, that in this c ase it is Defender , and not T estifier, who attacks Innoc ent. She is
the one who most dire ctly c auses there t o be c osts that others must choose ho w to distribute.
And y et although T estifier isn’t the attacker, he is not permitte d to let Innoc ent be harme d
b y Def ender. But neither is he permi tte d to re dire ct the c osts of the attack b ack onto De-
f ender — at least not if he has the option to instead re dire ct those c osts onto himself, as he
does in this c ase. After all: Def ender is just acting reasonab ly in light of the f alse inf ormation
f or which T estifier intentionally ga v e her. Def ender is acting e x actly as T estifier f oresa w and
intende d that she w ould. And T estifier w as under no pressure to giv e her this f alse inf orma-
tion. Giv en all of these c onsiderations, it se ems clear that T estifier has a duty to bear the c osts
of the attack, rather than let them f all on Innoc ent or re dire ct them onto Def ender. He is
morally re quire d to shield Innoc ent from the attack.
The important lesson f or present purposes is that a person c an ha v e duties to assume the
c osts of some one else’s attack. This point is the key to answ ering the Challenge of M erely
App arent Attackers. After all: merely app arent attackers like Jamm y and Bluff er are in much
91
the same position as T estifier. They , too , are highly responsib le and highly culp ab le f or in-
citing Def ender to attack some one, where this reaction is reasonab le in light of the evidenc e
they ha v e pro vide d her. (Here I agre e with F erzan that culp ability matters, and with M cM a-
han and Fro w e that responsibility matters. I just don’t think it matters in quite the w a y they
claim. M ore on this in §3.4.3.) The diff erenc e is just that they happen to also be the ta rget of
the attack. In the Jam and Bluff c ases, the person who incites the attack is also the target of
the attack. But it’s hard to se e wh y that diff erenc e should put them under an y less of a duty
to assume the c osts of Def ender’s attack than is T estifier.
Suppose that’s right. Suppose Jamm y and Bluff er ha v e a duty to assume the c osts of De-
f ender’s attack. Notic e that, on m y ac c ount of liability , this doesn’t estab lish that they are
liab le to be attacke d b y Def ender. After all, Jamm y and Bluff er don’t take on a duty to as-
sume the c osts of Def ender’s attack until after that attack has be en initiate d and there are
c osts that must f all on some one or other. And of c ourse w e c annot appeal to the go al of t hat
duty to justify the v ery actions that bring that duty into e xistenc e.
Inde e d, this is m y (unorthodo x) claim: Jamm y and Bluff er are not str ictly liab le to be at-
tacke d, sinc e attacking them does not serv e the go al of an y duty they ha v e prior to the attack.
Jamm y and Bluff er ha v e not f orf eite d rights against harm. But : they do f orf eit man y of the
perks or benefits that normally c ome with those rights. They bear man y of the upshots of
liability ev en though they are not strictly liab le.
20
They bear these upshots pre cisely be c ause they ha v e a duty to assume the c osts of De-
f ender’s attack. Let me e xp lain, one upshot at a time. Start with the first upshot of liability:
Jamm y and Bluff er w ould not be permitte d to fight b ack against Def ender with c omp arab le
20
This idea that the upshots of liability c an sometimes c ome ap art from liability will be f amiliar from chapter
tw o: §2.5.1.
92
f orc e.
This first upshot f o llo ws v ery straightf orw ardly from the claim that Jamm y and Bluff er
ha v e a duty to assume the c osts of Def ender’s attack. Fighting b ack w ould be a straightf or-
w ard w a y of f ailing this duty — it w ould be a w a y of re dire cting the c osts of the attack onto
Def ender.
Turn to the se c ond upshot: a b ystander w ould not be permitte d to fight b ack against
Def ender (with c omp arab le f orc e) on Jamm y or Bluff er’s behalf. This upshot is much less
straightf orw ard. It’s less straightf orw ard be c ause a b ystander, of c ourse, is not under a duty
to assume the c osts of Def ender’s attack.
But — and here’s the key — neither is Defender in this c ase re quire d to assume from
Jamm y and Bluff er the c osts of her attack. That ma y sound surprising giv en m y claim that
Jamm y and Bluff er are not strictly liab le to be attacke d. But refle ct f or a moment on the Hu-
man Shield c ase from abo v e. Suppose, in that c ase, that T estifier does what he is suppose d
to do: he thro ws himself in front of Innoc ent as a human shield. As he does so , Def ender
is made a w are of T estifier’s lies and Innoc ent’s innoc enc e. And no w suppose Def ender has
a choi c e. She c an allo w T estifier to suff er the bro ken leg in def ense of Innoc ent, or she c an
re dire ct the harm onto herse lf . Does Def ender ha v e a duty to do this? Does she ha v e a duty
to rescue T estifier from a bro ken leg at the c ost of a bro ken leg to herself?
Surely not. The reason Def ender lacks such a duty is that, ev en though she is the most
dire ct c ause of those c osts, and ev en though she is wrongfully targeting a non-liab le person
(Innoc ent), Def ender nonetheless bears less responsibility and culp ability f or those c osts than
T estifier. In virtue of this diff erenc e in responsibili ty and culp ability , T estifier has a duty to
assume those c osts from Def ender, but Def ender does not ha v e such a duty to w ards T esti-
93
fier.
21
No w c ome b ack to our characters, Jamm y and Bluff er. They are in an analogous position
with T estifier in this respe ct. They bear more responsibility and culp ability than Def ender f or
the c osts of her attack, and as such their duties of c ost assumption are as ymmetric.[f ootnoteI
mention both responsibility and culp ability be c ause they each make a c ontribution to Jamm y
and Bluff er’s duties to assume c osts.] Jamm y and Bluff er ha v e a duty to assume the c osts of
the attacker from Def ender, but not vic e v ersa.
This, then, is wh y a b ystander ma y not harm Def ender on Jamm y and Bluff er’s behalf. On
the ac c ount of liability I endorse, re c all, Def ender is liab le to be harme d b y b ystander only if
doing so w ould make Def ender to assume c osts she herself has a duty to assume. I’v e just
argue d that Def ender does not ha v e a duty to assume the c osts of her attack from Jamm y and
Bluff er. Thus, she is not liab le to be made , b y b ystander, to assume those c osts.
Turn finally to the third upshot of liability: Jamm y and Bluff er are not o w e d c ompensation
f or their injuries. The grounds of this third upshot are much the same as f or the se c ond. I’v e
just argue d that Def ender lacks a duty to assume the c osts of her attack from Jamm y and
Bluff er. But a duty to p a y c ompensation f or harming Jamm y and Bluff er w ould just be an ex
post v ersion of the duty to assume the c osts of her attack. P a ying c ompensation is just what
it loo ks like to assume the c osts of an attack after the damage has be en done. Sinc e Def ender
has no duty to assume the c osts of her attack, she has no duty to p a y c ompensation after the
attack has c onclude d.
21
It’s important to notic e here that duties of c ost-assumption are person-relativ e: y ou ma y ha v e a duty to
assume the c osts of an attack from X but not from Y . In the Human Shield c ase, f or instanc e, Def ender clearly
w ould be re quire d to assume c osts from In noc ent . Suppose Def ender c ould prev ent Innoc ent from being
harme d onl y b y assu ming t he c ost of a bro ken leg: she w ould be re quire d to do so . But if T estifier thro ws
himself in front of Innoc ent, Def ender does not likewise ha v e a duty to assume the c osts of a bro ken leg from
T estifi er .
94
T o summarize: characters like Jamm y and Bluff er are not strictly liab le to attacke d b y De-
f ender, in the sense that they entirely lack a right against such an attack. This is be c ause
attacking them does not serv e to make them assume c osts that they ha v e a duty to assume.
And that’s simp ly be c ause they pose no actual threat. Prior to the def ender’s reaction, there
are no c osts that ne e d distributing, and thus no c osts that Jamm y and Bluff er must be made
to assume. This is where m y ac c ount most drastic ally c omes ap art from Culpability and
Signaling . On the latter ac c ounts, Jamm y and Bluff er ha v e no right against attack. On
m y ac c ount, they do . And that right e x erts de ontic f orc e: this right c ontributes to making it
f act-relativ e impermissib le f or Def ender to attack them.
22
Jamm y and Bluff er do not f orf eit rights against harm. But they do f orf eit man y of the perks
that normally c ome with those rights. Normally , our rights against harm are enforc ea b le , in
the sense that those rights c an be def ende d with f orc e. And normally those rights are also
c ompensa b le , in the sense that their infringement demands c ompensation. Jamm y and Bluff er
do not f orf eit rights against harm, but they do render (some of) those rights non-enf orc eb le
and non-c ompensab le. They greatly w eaken those rights without entirely losing them. They
are liab le in (almost) all but name. They are — let’s sa y — quasi-lia b le .
23
W e’re finally in a position to se e m y answ er to the Challenge of M erely App arent Attack-
22
This, I think, is what e xp lains wh y w e c ould properly b lame Def ender f or attacking Jamm y and Bluff er if
she knew they pose d no actual threat. Jamm y and Bluff er ha v e a right not to be harme d whether or not Def ender
is awa re that they pose no actual threat; either w a y , Def ender lacks an obje ctiv e justific ation f or harming them.
But Def ender’s epistemic position does make a diff erenc e to her cu lpa bi lity f or attacking them. When she is not
in a position to kno w that her targets pose no threat, she has an e x cuse f or attacking them. But when she is in a
position to kno w , she lacks such an e x cuse. She w ould then cu lpa b ly do what is obje ctiv ely wrong.
23
There are a number of w a ys in which the orists ha v e previously note d that liability c an be “p artial ”. Almost
all the orists agre e that y our liability c an be p artial in the sense that y ou are liab le only to c ertain amounts of harm
(e.g., only to proportionate and ne c essary harm). Some the orists think liability c an be p artial in the sense that
y ou are liab le relativ e to some pe op le and not others (se e M cM ahan, “Limits of Self-Def ense”: 201-203; and m y
“Refusing Prote ction”: 46). M y idea in this p aper is that liability c an be p artial in y et another sense: one c an
enjo y some, but only some, of the usual “perks ” of ha ving a right.
95
ers. What distinguishes characters like Jamm y and Bluff er from characters like Innoc ent and
T win is that only the f ormer are quasi-liab le, in virtue of the f act that only the f ormer ha v e a
duty to assume the c osts of their def ender’s attack.
Jamm y and Bluff er ha v e such a duty , w e’v e said, be c ause they bear more responsibility and
culp ability than Def ender f or the f act that there are c osts that must be distribute d. But the
same is not true of Innoc ent and T win. They bear little or no responsibility f or that f act. And
they are in no w a y culp ab le f or that f act. That’s wh y they ha v e no duty to assume the c osts
of Def ender’s attack, and thus bear none of the upshots of liability .
3.4.3 Interlude: Responsibility and Culpability
Bef ore I turn to e xto l the virtues of this story , it might help to address a potential point of
c onfusion. I’v e claime d to be selling an answ er to the Challenge of M erely App arent Attack-
ers that is importantly diff erent from F erzan’s Culpability ac c ount and M cM ahan and
Fro w e’s Responsibility ac c ount. And y et the notions of responsibility and culp ability
p la y a c entral ro le in m y answ er. Whenc e the diff erenc e?
Thre e brief c omments. First, one point of disagre ement has to do with the oretic al inclu-
sivity . I think responsibility and culp ability are bot h important elements of the full story ,
and that it’s a mistake to ha v e a the ory of liability that has a ro le f or one but not the other.
A f amiliar reason to not f ocus e x clusiv ely on culp ability c omes from c ases in v o lving e x cuse d
attackers — e.g., a c ase like F alse T estimon y . Def ender wrongfully , but non-culp ab ly , attacks
Innoc ent. Nonetheless, Def ender’s responsibility f or the attack puts her under a duty to as-
sume those c osts from Innoc ent, thereb y making her liab le to def ensiv e harm from Innoc ent.
But there are also reasons to not f ocus e x clusiv ely on responsibility at the e xpense of culp a-
96
bility . One reason is that there are c ases where the person who is most responsib le f or a state
of aff airs is diff erent from the person who is most culp ab le f or that state of aff airs. Here’s an
e x amp le:
Omissi on . Def ender has be en giv en f alse, but c ompelling, testimonial evidenc e
that Innoc ent is an imminent, wrongful threat. Def ender ro lls a boulder to-
w ards Innoc ent, believing this is the only w a y to def end herself. V illain is nearb y ,
and witnesses the sc ene. He c an c ostlessly div ert the boulder such that no one
will be harme d. V illain hates Innoc ent, ho w ev er, and w ould be glad to se e her
kille d. He refrains from re dire cting the boulder. F ortunate f or Innoc ent, she
c an def end herself either b y re dire cting the boulder to w ards Def ender or V il-
lain.
I think that Def ender is most responsi b le f or the threat to Innoc ent.
24
But V illain is clearly
most cu lpa b le f or that threat. And it is clearly on him, rather than Def ender, that the c osts
should f all. This c ase sho ws that culp ability makes a c ontribution to liability abo v e and
bey ond its imp lic ations f or responsibility . On m y view , that’s be c ause culp ability c an c on-
tribute to the strength of a person’s duties of c ost assumption abo v e and bey ond its imp lic a-
tions f or responsibility .
Se c ond, the Assumption ac c ount insists that responsibility and culp ability c ontribute
to (quasi-) liability only when and be c ause they produc e a duty of c ost-assumption. They
24
Reasons f or thinking Def ender the more responsib le p arty include the f acts that (i) Def ender makes a
greater c ausal c ontribution to the threat to Innoc ent than V illain; (ii) Def ender’s c ontribution to that threat
is no less intentional than V illain’s c ontribution-b y-omission to that threat; and (iii) Def ender is doing harm
where V illain is only allo wing harm. I take responsibility to be a morally-neutral c onc ept that f ocuses on the
degre e of c ontro l or “authorship” that an agent e x ercises o v er an outc ome. Def ender e xhibits more c ontro l o v er
the threat to Innoc ent than does V illain, and is more aptly desc ribe d as the “author” of that threat.
97
only w ork through this interme diary . This ma y se em like a c omp atib le addendum to Re-
sponsibility and Culpability . But I don’t think it is. This is be c ause — as w e’ll se e in
the ne xt se ction — duties of c ost assumption admit of a p lurality of grounds other than f acts
about responsibility and culp ability . This means that on Assumption , responsibility and
culp ability are neither individually ne c essary nor joi nt ly ne c essary f or (quasi-)liability .
Third, as w e’v e alre ady se en, I p art w a ys with Responsibility and Culpability in that
I think responsibility / culp ability f or posing an actua l threat does not ha v e the same kinds
of liability eff e cts as responsibility / culp ability f or a ppea ri ng to pose a threat. When some-
one poses an actual, wrongful threat, they’v e already c reate d the c osts that ne e d distributing,
and their responsibility / culp ability f or those c osts c ontributes to putting them under a duty
to assume the c osts. That makes them strictly liab le to be harme d in servic e of that duty .
But when some one m erely appears to pose a threat, there are not (y et) c osts than ne e d dis-
tributing. They are thus not (y et) u nder a duty to assume an y c osts, and so their responsibil-
ity / culp ability f or appearing to pose a threat does not giv e rise to an y liability . It’s only onc e
the def ender reacts and generates the ne e d to assume c osts that the responsibility / culp ability
f acts generate the duty of c ost assumption and the quasi-liability that I’v e claime d f o llo ws.
3.5 Advantages
The Assumption ac c ount bef ore us, let me no w turn to sa y wh y I f a v or the Assumption
ac c ount, ap art from the f act that I think it is an upshot of p lausib le ideas about the relation-
ship betw e en duty and liability and about the w a ys in which w e c an take on duties to assume
c osts.
By f ar the biggest selling point of the the ory is its e xp lanatory reach: it e xp lains a wider
98
range of p henomena than either Culpability or Signaling . In this se ction I’ll mention
tw o important respe cts in which this is true.
3.5.1 The Challenge of No-Threat Liability
First, Assumption answ ers the Challenge of M erely App arent Attackers while also answ er-
ing a more general challenge — what I’ll c all the Challenge of No-Threat Lia bi lity .
Here’s the thing: c ases of merely app arent attackers are just one kind of c ase where persons
who pose no actual threat bear the upshots of liability . There are other c ases. The Challenge
of M erely App arent Attackers is just one p art of a more general challenge.
I’ll giv e tw o e x amp les in which a person bears the upshots of liability despite neither pos-
ing, nor ev en a ppea ri ng to pose , an y actual threat. The first is an e x amp le of what Sab a Bazargan-
F orw ard c alls “ c omp licitous liability”:
25
Getawa y Driver . V illain w ants to se e Def ender kille d. So he hires an assassin
to do the job . After sending Assassin on his w a y , V illain hires some one else to
serv e as Assassin’s geta w a y driv er without Assassin’s kno wing. The driv er w aits
on the stre et as Assassin thro ws a grenade into Def ender’s home. The grenade
will break Def ender’s leg unless she kicks it to w ards Driv er bef ore it detonates.
Def ender kic ks the grenade to w ards Driv er, breaking his leg.
W e c an suppose that Driv er has not made an y c ausal c ontribution (ev en b y omission) to the
threat pose d b y Assassin. After all, he makes no diff erenc e to Assassin’s actions, espe cially as
the latter doesn’t ev en kno w that Driv er is present. Driv er kno ws full w ell that he is signing
up to be a geta w a y driv er f or an assassination, but he does not p la y a ro le in the p lanning
25
Bazargan, “Comp licitous Liability in W ar,” Phi losop hi ca l Stu di es 165 (2013): 177-195.
99
or e x e cution of the assassination. His ro le is limite d to being ready to c ontribute after the
assassination.
Giv en these stipulations, w e c an hardly sa y that Driv er is responsible or cu lpa b le f or the
threat to def end. He is, of c ourse, culp ab le f or somet hi ng : he is culp ab le f or choosing to be
Assassin’s geta w a y driv er. But he isn’t culp ab le f or the t hreat to Defender , sinc e he makes no
c ontribution to that threat. As such, both the Responsibility and Culpability ac c ounts
pre dict that Driv er is not liab le to ha v e the grenade kicke d his w a y .
W e also c an hardly sa y that Driv er perf orms an y beha vior that c on v entionally signals ag-
gression to w ards Def ender. He doesn’t. And so Signaling likewise pre dicts that Driv er is
not liab le to ha v e the grenade kicke d his w a y .
Driv er does not pose a threat. He does not c ontribute to the threat that some one else
poses. He does not ev en appear to do either of these things. And y et he sure se ems liab le to
be harme d. He bears the upshots of liability . F or e x amp le: it w ould be wrong f or him to kick
the grenade bac k to w ards Def ender. And the reason surely has to do with the f act that he is
c omp li cit in the assassination, ev en though he has not c ontribute d to it.
It’s an intuitiv e idea that c omp licity c an ground liability . One of the virtues of Assump-
tion is that it helps us to make sense of this idea. This is be c ause it is independently p lausib le
that c omp licity c an ground a duty of c ost assumption. W e c an se e this most clearly in the ex
post c onte xt. Imagine, if y ou will, that Assassin suc c e e de d in breaking Def ender’s leg and that
Driv er helps him to esc ape as p lanne d. Assassin and Driv er, ho w ev er, are later c onfronte d b y
the po lic e: Assassin dies in a shootout and Driv er is taken into custody . Assassin — being
dead — is unab le to c ompensate Def ender f or her injuries. Does Driv er then ha v e a duty to
p a y c ompensation? It sure se ems that he does. But if he has a duty to assume the c osts ex post ,
100
it’s hard to se e wh y he w ouldn’t ha v e such a duty ex a nte .
So Assumption deliv ers the intuitiv e v erdict that Driv er bears some liability to ha v e the
grenade kicke d his w a y . By kicking the grenade his w a y , Def ender is just making him to as-
sume c osts he has a duty to assume.
26
Here’s a se c ond c ase where Assumption deliv ers where c ompeting the ories c annot:
Po li c e Shi e ld . A po lic e offic er, P atrick, takes an o ath upon entering the servic e
— an o ath to prote ct the citizens of his to wn ev en at a risk of serious harm to
himself. While on p atro l one da y , he witnesses a gunman enters a pub lic p ark.
The gunman points a gun at an innoc ent civilian and prep ares to shoot at one
of her legs. P atrick is not in a position to disarm the gunman, but he is in a
position to thro w himself betw e en the gunman and the civilian, and take the
harm of a bro ken leg upon himself. P atrick, ho w ev er, chooses not to shield the
civilian. But, as it happens, the civilian has tw o w a ys to def end herse lf . She has
a metal she et, which she c an use either to defle ct the bullet a w a y from herself
and to w ards P atrick (breaking his leg) or to defle ct the bullet to w ards another
innoc ent b ystander (breaking their leg).
There’s a lot w e might ask about this c ase. But I w ant to f ocus on one f eature in p articular,
which is that if the civilian is going to use the shield, she should defle ct the bullet to w ards
P atrick rather than to w ards the b ystander. There is at least more justific ation f or harming
P atrick than f or harming the b ystander.
26
Notic e tw o things about this c ase. First, Driv er is liab le in the strict sense; he’s not just quasi-liab le. By
kicking the grenade his w a y , Def ender is making Driv er to assume the c osts of an attack that temporally pre c e des
her o wn use of f orc e. Se c ond, although Driv er ma y not kick the grenade b ack to w ards Def ender, he w ould be
permitte d to kick it to w ards Assassin. This is be c ause, a lthough Driv er has a duty to assume the c osts of the
attack from Defender , he w ould clearly not ha v e such a du ty with respe ct to Assassin.
101
Assumption off ers an e xp lanation. On this ac c ount, P atrick bears some liability to ha v e
the bullet dire cte d his w a y . Not be c ause he poses an y threat; not be c ause he appears to
pose an y threat; not be c ause he is c omp licit in the attack. Rather, he’s liab le be c ause he has
promise d to risk lif e and limb to def end Civilian. Promises are just another w a y in which du-
ties of c ost assumption might c ome about. P atrick’s promissory duty to assume c osts from
Civilian e xp lains wh y there’s more reason to defle ct the bullet to w ards him than to w ards the
b ystander.
27
By c ontrast, Culpability , Responsibility , and Signaling don’t help to e xp lain this.
P atrick is not culp ab le or responsib le f or the gunman’s threat, and he’s perf orme d no signal
of aggression. Culpability , Responsibility , and Signaling thus pre dict no diff erenc e
in liability betw e en P atrick and the b ystander. And if both persons are e qually non-liab le,
then it’s hard to se e what other c onsiderations c ould f a v or harming P atrick o v er harming the
b ystander in pre cisely the same manner.
The Assumption ac c ount giv es us a more c omp lete answ er to the general Challenge of
No- Threat Liability . But I do think the f oregoing discussion helps us to better appre ciate
what Culpability , Responsibility and Signaling ha v e to teach us about the c ondi-
tions of liability . These ac c ount are instructiv e. They are instructiv e in that they help us to
understand some of the w a ys in which duties of assumption c an arise. Responsibility and/ or
culp ability f or appearing to bear a threat c an lead to liability pre cisely be c ause it c an lead to a
duty of c ost assumption. Likewise, signals of aggression c an lead to liability pre cisely be c ause
they t oo c an lead to a duty of c ost assumption.
27
E v en if P atrick’s promissory duty does not make him liab le to t hat mu c h harm, his liability w ould still c on-
tribute to a “mix e d justific ation” f or harming him, which w ould still e xp lain wh y Civilian has more justific ation
f or harming P atrick than the b ystander.
102
These ac c ounts f all short, ho w ev er, in that they latch on to only some of the w a ys w e c an
incur such duties. As c ases like Geta w a y Driv er and P o lic e Shield illustrate, the re are ot her
w a ys. And thus, there are other w a ys w e c an be c ome liab le to harm.
28
3.5.2 The Challenge of Necessity
Another adv antage of Assumption is that it helps us to make better sense of the so-c alle d
“ne c essity c onstraint” on def ensiv e harm.
Consider a p aradigm c ase of unne c essary harm in an actual-threat c ase:
Easy Defense . A villain attempts to murder Def ender. Def ender kno ws that she
has tw o e qually good w a ys to disarm V illain: b y killing him or b y giving him a
light s lap on the wrist.
It’s widely held that it is wrong f or Def ender to kill V illain under these circumstanc es.
It’s wrong be c ause it isn’t “ne c essary” f or Def ender to kill the attacker to def end herself, and
attackers (like V illain) make themselv es liab le only to ne c essary def ensiv e harm.
29
This intuitiv e idea raises an ob vious challenge, ho w ev er. Consider characters like Jamm y
and Bluff er. They pose no actual threat. And so of c ourse it isn’t ne c essary f or Def ender
to harm them in order to def end herself. The challenge, then, is this: w e ne e d to e xp lain
wh y characters like Jamm y and Bluff er bear the upshots of liability without undermining
28
Cases in v o lving “hidden attempters ” are another kind of c ase that makes troub le f or Culpability , Re-
sponsibility and Signaling . The T w o Snipers Re dux c ase in §3.5.2 is an e x amp le of such a c ase.
29
Numerous the orists endorse this idea that the prob lem with unne c essary def ensiv e harm is that it wrongs
the attacker. Se e, f or e x amp le, Kaila Draper, “Ne c essity and Proportionality in Def ense,” in The Et hi cs of
Se lf-Defense , e ds. Christian Coons and Michael W eber (Oxf ord: Oxf ord Univ ersity Press, 2016): 174-175;
M cM ahan, “Limits of Self-Def ense,” 195-197; and Quong, Defensive Forc e : 129-132. But se e Fro w e, Defen-
sive Ki l li ng : 88-119, f or a dissenting opinion. F or a response to (some of) Fro w e’s arguments, se e M cM ahan,
“Limits of Self-Def ense,” 195-206.
103
our ability to e xp lain wh y w e shouldn’t kill V illain in a c ase like E as y Def ense. Call this the
Cha l lenge of Nec essity .
Assumption has a simp le answ er to this challenge. On this ac c ount, re c all, although
Jamm y and Bluff er bear the upshots of liability , they are not strictly liab le to be attacke d.
They are not liab le to be attacke d be c ause attacking them does not serv e an y duty that they’re
under. It’s only after the attack has begun that they take on a duty to assume the c osts of the
attack. Bef ore that, there aren’t an y c osts that ne e d distributing. This is a c rucial diff erenc e
betw e en merely app arent attackers and actual attackers. Actual attackers f orf eit rights against
harm. M erely app arent attackers like Jamm y and Bluff er only f orf eit some of the perks that
c ome with those rights. What they lose, as w e’v e se en, is the permission that they and others
normally ha v e to def end those rights, as w ell as their claim to c ompensation f or the infringe-
ment of those rights. They greatly w eaken their rights without entirely losing them.
So what Assumption giv es us is an e xp lanation f or wh y Jamm y and Bluff er bear the up-
shots of liability that is perf e ctly c omp atib le with the princip le that attackers are only liab le
to ne c essary def ensiv e harm.
By c ontrast, Culpability , Responsibility , and Signaling both attribute full-b loode d
liability to Jamm y and Bluff er. And so these ac c ounts are both inc onsistent with the idea that
liability is restricte d to ne c essary harm. T o answ er the Challenge of Ne c essity , then, these ac-
c ounts ne e d some other princip le b y which to e xp lain wh y Def ender ma y not kill V illain in a
c ase like E as y Def ense.
No w y ou might think that it’s not so hard to find such a princip le. Y ou might think that
the prob lem with killing V illain isn’t that it isn’t ne c essary , but rather that it doesn’t a ppea r
ne c essary . That is, w e might e xp lain the c ase of E as y Def ense b y appeal to the princip le that
104
lia bi lity is restri cted to ha rm t hat a ppea rs nec essa ry . This ref ormulation of the ne c essity c on-
straint is c onsistent with the thought that Jamm y and Bluff er are liab le to harm, sinc e harming
them does at least appear to be ne c essary .
But that princip le it p lainly f alse, as w e c an se e from c ases of hidden attackers. Consider:
Two Sni pers . There are tw o snipers hidden in a bellto w er, each attempting to
kill Def ender. Def ender only spots Sniper
1
, and believ es he is the only person in
the bellto w er. She kno ws that she c an sa v e her lif e only if she thro ws a grenade
into the bellto w er (she doesn’t kno w that this is also ne c essary to def end against
Sniper
2
. She thro ws the grenade, w ounding both snipers.
Sniper
2
is p lainly liab le to this harm, ev en though it doesn’t appear to be ne c essary to De-
f ender to harm him. So w e sho uld reje ct the princip le that lia bi lity is restri cted to ha rm t hat
a p pea rs nec essa ry .
W e might instead try the f o llo wing w eaker, disjunctiv e princip le: lia bi lity is restri cted to
ha rm t hat eit her is nec essa ry or t hat a ppea rs to t he defender to be nec essa ry . But this princip le
also runs into troub le in c ases like
Two Sni pers Redux . As bef ore, there are tw o snipers hidden in a bellto w er, each
attempting to kill Def ender. Def ender only spots Sniper
1
, and believ es he is the
only person in the bellto w er. She kno ws that she c an sa v e her lif e only if she
disarms him. She c an do so b y thro wing a grenade into one of tw o windo ws.
If she thro ws the grenade into the first windo w it will break the legs of Sniper
1
and Sniper
2
. If she thro ws the grenade into the se c ond windo w it will break the
legs only of Sniper
1
. As Def ender is una w are of the presenc e of Sniper
2
, she se es
105
no reason to pref er one windo w to the other. She chooses to thro w the grenade
into the first windo w . Both snipers are w ounde d. A s it turns out, Sniper
2
’s gun
w as jamme d and he pose d no actual threat.
Sniper
2
surely bears the upshots of liability . He w ould not be permitte d to kick the grenade
b ack to w ards Def ender, b ystanders w ould not be permitte d to kick the grenade b ack on
his behalf, and he is not o w e d c ompensation f or his injuries. Ac c ounts that appeal to full-
b loode d liability to e xp lain the marks of liability will w ant to attribute liability here, then. But
notic e: it is neither ne c essary f or Def ender to harm Sniper
2
(the option to thro w the grenade
into the other windo w w as a v ailab le to her), nor does it a ppea r to Def ender to be ne c essary
to harm him (she doesn’t believ e that he’s a threat; she doesn’t ev en kno w he’s there). The
disjunctiv e princip le also se ems f alse.
No w perhaps there is y et some w a y f or these ac c ounts to ha v e their c ake and eat it too .
P erhaps there’s some w a y to insist on strict liability f or characters like Jamm y and Bluff er
without undermining our ability to e xp lain what’s wrong with killing V illain in a c ase like
E as y Def ense. I don’t take m yself to ha v e giv en an e xhaustiv e proof that there isn’t a w a y to
strike this b alanc e. There ma y y et be some alternativ e princip le I’v e misse d that c an do the
job . What I hope to ha v e sho wn, though, is that there is a real puzzle here that puts a burden
on a c c ounts that attribute full-b loode d liability to merely app arent attackers.
It’s an attractiv e f eature of Assumption that it bears no such burden. It doesn’t re quire
us to giv e up the intuitiv e idea that liability is c onstraine d b y c onsiderations of ne c essity .
And in f act there’s another attractiv e f eature of the ac c ount here, which is that it helps us
to make sense of the f act that it is sometimes permissib le to resist unne c essary harm and it
sometimes is not. T o se e this, c omp are tw o c ases of unne c essary harm:
106
Irresisti b le Defense . V illain is fully culp ab le f or attempting t o kill Def ender. De-
f ender c an sa v e her lif e either b y breaking one or both of V illain’s leg. But De-
f ender believ es she c an def end herself only b y breaking both of V illain’s legs.
And so she attempts to break both his legs.
Resisti b le Defense . Driv er is in the mood f or pizza. He kno wingly risks driving
on ic y ro ads to pick one up , and loses c ontro l of his v ehicle. It c are ens to w ards
Def ender, who is sitting on her front porch. She kno ws that she c an just as
w ell prev ent the c ar from hitting and killing her either b y breaking one or both
of Driv er’s legs. But Def ender doesn’t like Driv er v ery much, and so takes this
opportunity to attempt to break both his legs.
In either c ase, Def ender chooses a def ensiv e option that i s unne c essary . But there se ems an
important diff erenc e. F or one thing, it se ems that Driv er w ould be permitte d to use at least
modest amounts of f orc e to fight b ack against Def ender. Suppose, f or e x amp le, that he had a
w a y to re distribute the harm such that he and Def ender each suff er a bro ken leg, rather than
him suff ering tw o bro ken legs. I think he w ould be permitte d to do this. But the same is not
true of V illain. He w ould not be permitte d to resist in such a w a y .
W e c an makes sense of this diff erenc e b y appeal to the idea at the heart of m y answ er to the
Challenge of M erely App arent Attackers — the idea that a person c an ha v e a duty to assume
the c osts of someone e lse ’s wrongful attack. V illain is not permitte d to fight b ack be c ause t hat
permission is undercut b y his duty to assume the c osts of Def ender’s mistake — namely , the
c ost of an e xtra, unne c essary bro ken leg. It’s independently p lausib le that V illain has such a
duty . And he has such a duty , I think, f or reasons of c omp arativ e culp ability . Def ender makes
a mistake, y es, but her mistake is at least a reasonab le one in light of her evidenc e. It’s a non-
107
culp ab le mistake. V illain, b y c ontrast, b ears a great deal of culp ability f or putting Def ender
in a position to make that mistake. V illain’s duty to assume the c osts of the wrongful harm
undercuts the permission he w ould otherwise ha v e to re dire ct those c osts onto Def ender.
Things are diff erent f or Driv er. He bears minimal culp ability f or his threat to Def ender.
And what’s more: in his c ase, Def ender is at f ault f or the unne c essary harm she attempts to
impose on Driv er. She kno ws better. Under these c onditions I find it v ery p lausib le that
Driv er (unlike V illain) is not under a duty to assume the c osts of Def ender’s mistake.
That’s wh y Driv er c an resist and V illain c annot. The diff erenc e isn’t that the one is liab le
to unne c essary harm and the other isn’t. The diff erenc e is that only one has a duty to assume
harm t hat he w as not liab le to ha v e presse d against him in the first p lac e.
3.6 Conclusion
Let’s take stock of where w e’v e c ome. There’s a challenge f or the the ory of liability , the Chal-
lenge of M erely App arent Attackers. I’v e argue d that the Assumption ac c ount off ers the
resourc es to answ er this challenge. Assumption claims that a person is liab le to be made to
bear harm they ha v en’t c onsente d to when and be c ause they’ d be re quire d to take that harm
upon themselv es. On this ac c ount, merely app arent attackers like Jamm y and Bluff er are not
strictly liab le to harm, but they nonetheless bear most of the upshots of liability . They are
quasi-lia b le .
M ore o v er, I’v e argue d that this answ er to this challenge also helps us answ er the more gen-
eral Challenge of No- Threat Liability as w ell as the Challenge of Ne c essity . It ev en helps us
to understand wh y some unne c essary def ensiv e harms c an be resiste d and wh y some c an’t be.
A closing thought. As I’v e presente d it here, Assumption is inc omp lete. I’v e claime d
108
that duties of assumption are the grounds of liability and quasi-liability . I’v e claime d that
there are a v ariety of w a ys such duties might arise, and I’v e pointe d to a f ew such w a ys. F or
e x amp le, I’v e claime d that a person c an ha v e a duty to assume c osts in virtue of their duty not
to harm, or in virtue of their responsibility f or those c osts, their culp ability f or those c osts,
their c omp licity in those c osts, or be c ause they ha v e made c ertain promises. But this is where
there is a great deal more to be said. This is a research proje ct that c an’t be tackle d in a single
p aper. Filling out Assumption re quires c ataloguing these diff erent grounds of assumptiv e
duties and of understanding ho w they interact with one another.
What I hope to ha v e ac c omp lishe d in the present p aper, ho w ev er, is to motiv ate the idea
that t his where much important action lies in the the ory of def ense. I hope to ha v e sho wn that
the w a y to make progress in understanding liability ac ross the full range of c ases is to make
progress in understanding when and wh y w e ha v e duties to assume c osts from one another.
109
4
Chapter F our: The Pric e of Duty
4.1 Introduction
State officials regularly impose harm in the line of duty , some of it wrongful.
1
F or some of
these wrongful harms, ho w ev er, it is widely believ e d that state officials should not ha v e to p a y
1
By ‘wrongful harm’ I mean harm that is rights-infringing. Be c ause rights ma y sometimes be permissib ly
infringe d, a wrongful harm (on this w a y of using the term) ne e dn’t be morally impermissib le. Likewise, the
agent of wrongful harm ne e dn’t be b lamew orth y .
110
c ompensation. Consider, f or e x amp le:
Fi re . Citizen’s house is hit b y lightning and erupts in flames. Citizen is trappe d
inside. T o sa v e Citizen, Firefighter b lasts the house with w ater, e xtinguishing
the flames. As an unintende d and una v oidab le side-eff e ct, she c auses thousands
of do llars of w ater damage to Neighbor’s home.
Mista ken Imprison ment . Judge e x ercises great c are in his handling of a c riminal
c ase, and ac curately app lies the la w in light of his a v ailab le evidenc e. He sen-
tenc es V ictim to prison. Guard transports V ictim to prison and spends man y
y ears prev enting V ictim from esc aping — until new evidenc e emerges pro ving
that V ictim w as innoc ent all along.
Neighbor and V ictim suff er wrongful harms f or which they are o w e d c ompensation. But it
se ems clear that Firefighter should not ha v e to be the one to c ompensate Neighbor. Likewise,
it se ems clear that neither Judge nor Guard should ha v e to be the one to c ompensate V ictim.
Compensation should be p aid — perhaps b y the state, perhaps b y the citizenry — but not
b y these individual officials.
Let’s giv e the general p henomenon a name. Not to be c onfuse d with the legal p henomenon
of the same name, w e’ll c all it
Official Immunity . State officials (morally) should not ha v e to bear the c osts
f or some of t he wrongful harms they impose in the line of duty .
This w a y of desc ribing the p henomenon is meant to be e cumenic al betw e en v arious f orms
this immunity might take. It might be, f or e x amp le, that Firefighter’s immunity c onsists in
111
the simp le absenc e of a moral duty to c ompensate Neighbor. Or it might be that Firefighter
does ha v e a pro tanto duty to c ompensate Neighbor, but that she a lso has a right of i ndem-
nifi cati on — that is, a right that some third-p arty (such as the state) assume the c osts of her
c ompensatory duty . As I’ll use terms, either of these w ould c onstitute a f orm of official im-
munity .
The present chapter dev elops a the ory of the grounds, nature, and limits of official im-
munity . Wh y do state officials enjo y such immunity? What f orm does this immunity take?
What are the sorts of harms f or which state officials do (not) enjo y immunity?
The chapter proc e e ds as f o llo ws. In §4.2 I c onsider and reje ct v arious w a ys of e xp laining
official immunity b y appeal to the princip al-agent relationship that ho lds betw e en the state
and its officials. In §4.3 I turn from the relationship betw e en the state and its officials to
loo k at the relationship betw e en the state and the citizenry . I c onsider and reje ct the idea that
official immunity is best e xp laine d b y appeal to the f act that citizens often benefit from the
actions of state officials. In §4.4 I present m y positiv e ac c ount — the Price of Duty ac c ount
— ac c ording to which official immunity is e xp laine d b y the spe cial duties of c ost sharing
that f all on persons f or whom p articip ation in a c ooperativ e proje ct is morally re quire d. In
§4.5 I sho w ho w this ac c ount helps us make progress in understanding the limits of official
immunity . Spe cific ally , it helps us identify f our types of damages f or which state officials are
not immune. I c onclude in §4.6 b y noting the justific atory burden this ac c ount puts on the
regime of legal immunity that w e find, f or e x amp le, in the Unite d States.
112
4.2 The Principal-Agent Relation
Here is a notab le f eature of man y c ases of official immunity: a priv ate citizen acting on their
o wn wou ld be liab le f or man y of the sorts of wrongful harms f or which state officials are
immune. Consider, f or e x amp le, a v ariant of the Fire c ase:
Fi re 2 . Citizen’s house is hit b y lightning and erupts in flames. There isn’t time
to w ait f or the fire dep artment, and Citizen c an sa v e himself only b y putting the
fire out with his garden hose. As an unintende d and una v oidab le side-eff e ct, he
c auses tho usands of do llars of w ater damage to Neighbor’s home.
Whereas it se ems Firefighter should not ha v e to c ompensate Neighbor in the original Fire
c ase, it se ems that Citizen s hou ld ha v e to c ompensate Neighbor in Fire 2. What’s the rele-
v ant diff erenc e? One c andidate that jumps out is the f act that Firefighter acts as an agent
of another (the state or the citizenry) whereas Citizen acts under his o wn authority . This
diff erenc e is espe cially notab le in light of the f act that the princip al-agent relationship has
be en long re c ognize d as a sourc e of immunity in priv ate la w in man y nations. If, f or e x amp le,
an emp lo y e e of trucking c omp an y A damages some of the c argo she is hauling f or c omp an y
B, the emp lo y e e ma y enjo y a legal right to indemnific ation against her emp lo y er. That is,
although the emp lo y e e w ould ha v e a legal duty to c ompensate c omp an y B f or the damages
she c ause d, c omp an y A w ould ha v e a legal duty to c o v er those c ompensatory c osts.
2
(This
isn’t the sort of immunity that c onsists in the simp le absenc e of duty; rather it’s the sort of
immunity that c onsists in a right that some one else bear the c osts of one’s duty — a right of
indemnific ation.)
2
Americ an La w Institute, Restatement (Thi rd) of Agency (St. P aul, Minnesota: Americ an La w Institute
Pub lishers, 2006), ch. 7.
113
No w it might be that there is no justific ation f or this f eature of priv ate la w . Or it might be
that it admits only of an instrumentalist justific ation: perhaps this po lic y does not serv e to
enf orc e an y ante c e dent moral duties that emp lo y ers ha v e to indemnify their emp lo y e es, but
is justifie d only b y , f or e x amp le, its distributiv e or inc entiv e eff e cts.
3
P erhaps. In this se ction, ho w ev er, I w ant to take seriously the idea that there is a de eper
moral justific ation behind priv ate agenc y la w . I w ant to take seriously the idea that agents
ha v e an ante c e dent moral right to be indemnifie d b y their princip als f or c ertain c osts they
c ause in their c ap acity as agents. I c onsider tw o stories that might be to ld about wh y agents
ha v e such rights. I argue, ho w ev er, that these e xp lanations c annot p lausib le e xtend to ac c ount
f or the moral immunity enjo y e d b y state officials.
4.2.1 The Promissory Account
One story claims that princip als ha v e a moral duty to indemnify their agents be c ause they
ha v e promised to do so . In the c ase of priv ate emp lo y ers, f or e x amp le, it might be thought
that they make a promise to indemnify their emp lo y e es — perhaps not e xp licitly , but at least
imp licitly , in virtue of the f act that it is c on v entionally understood that emp lo y ers assume
a duty of indemnific ation when they hire some one. P erhaps the same is t rue of state offi-
cials: they ha v e a right to be indemnifie d b y either the state or the citizenry when and be c ause
the state or the citizenry has promise d to indemnify them. Call this the Pr o missor y A c-
count . On this ac c ount, promises are the grounds of official immunity .
3
A distributiv e argument f or this po lic y might appeal to the f act that emp lo y ers are typic ally better po-
sitione d than their emp lo y e es to spread c osts (to c onsumers, insurers, and the like). An inc entiv e argument
might appeal to the f act that emp lo y ers — ha ving the ability to make organizational changes — are typic ally bet-
ter positione d than their emp lo y e es to re duc e future emp lo yment-relate d damages. Sinc e an indemnific ation
re quirement w ould inc entivize emp lo y ers to make such changes, total future damages w ould be minimize d.
114
Whether or not promises e xp lain the immunity enjo y e d b y priv ate emp lo y e es, w e should
reje ct the Promissory Account . Consider, first, the idea that state officials are immune
when and be c ause the state has promise d to indemnify them. The prob lem with this idea
is that official immunity is not p lausib ly c ontingent on whether the state has promise d to
indemnify its officials. Imagine a state that has not promise d to indemnify its officials. That
is, imagine a state whose la w does not besto w on the state a legal duty to indemnify its officials.
But no w c onsider our p aradigm c ases of official immunity against the b ackdrop of such a
state: Firefighter c auses damage to Neighbor’s home; Judge and Guard mistakenly imprison
the innoc ent V ictim. Should these c osts f all on Firefighter, Judge, or Guard in a w orld where
the la w is silent about indemnific ation? Surely not. Altering the legal b ackdrop in this w a y
does not change the intuitiv e judgment that Firefighter, Judge, and Guard should not ha v e
to bear the c osts of c ompensating Neighbor or V ictim.
The same prob lem ho lds f or the idea that state officials are immune when and be c ause the
citi zen ry has promi se d to indemnify them. And here w e ne e dn’t ev en use our imaginations:
it isn’t p lausib le that most citizens in rea l-world po litic al c ommunities promise to indemnify
their officials. Promises are c omp lic ate d c reatures, but the f o llo wing is a p lausib le minimal
c onstraint on the making of a promise:
Epistemic Constrain t . X promises to φ only if X perf orms some action A such
that (i) X believ es or intends that A c onstitutes a promise to ψ (where ψ -ing is
or entails φ -ing), or (ii) it is pub lic kno wle dge that A is a type of action that
c onstitutes a promise to ψ .
This princip le is e xtremely p lausib le. Promises alter the moral boundaries of the promisor:
they alter their duties and liabilities. Sinc e it is in our interest to ha v e c ontro l o v er such
115
changes to our moral boundaries, it is in our interest to ha v e a good amount of c ontro l o v er
when w e make promises. Likewise, be c ause promises also change the moral boundaries of the
promise e (who takes on new rights) and be c ause social c oordination often re quires us to un-
derstand our moral relations to one another, it is also in our interest to be ab le to kno w when
ot hers make promises. But w e’ d ha v e neither much c ontro l o v er when w e make promises nor
find it eas y to kno w when others make promises if Epistemic Constraint w ere f alse.
4
This p lausib le princip le makes serious troub le f or the idea that most citizens ha v e promise d
to indemnify their state officials. If aske d, most citizens w ould surely assert that they ha v e
nev er promise d such a thing. This is good evidenc e, not just that those citizens ha v en’t done
something they believ e d or intende d to c onstitute a promise to indemnify , but that they
ha v en’t ev en done something that is pub licly kno wn to c onstitute a promise to indemnify .
It is v ery imp lausib le that it c ould be pub lic kno wle dge among the citizenry that φ -ing c on-
stitutes a promise to indemnify state officials but that most citizens ha v e done φ without
believing that they ha v e promise d to indemnify state officials.
4.2.2 The Responsibility Account
The Promissory Account isn’t promising. But here’s a diff erent w a y one might think
to e xp lain official immunity in terms of the princip al-agent relation. P erhaps the princip al-
agent relationship is the sourc e of official immunity , not be c ause it in v o lv es c ertain promises,
but be c ause of its eff e ct on the distribution of responsibility .
The idea is this. Compensatory duty tends to track responsibility . The p arty who typi-
4
F or an e x c ellent discussion along these lines, se e René e Jorgensen, “M oral Risk and Communic ating Con-
sent,” Phi losop hy a nd Pu b li c Aff a i rs 47 (2019): 179-207. Jorgensen is f ocuse d on c onsent, rather than promise,
but much of her discussion app lies to both p henomena.
116
c ally o w es the most c ompensation f or some harm is the p arty who w as most responsib le f or
producing the harm.
5
But agents are i nstru ments of their princip als. As such, when agents
c ause wrongful harm in their c ap acity as agents, it is the princip al who is the “first c ause”,
who is ultimately in c ontro l, and who is thus the primary locus of responsibility f or (man y
of) the wrongful harms produc e d b y the agent. Be c ause c ompensatory duty tracks respon-
sibility , then, it is the princip al, rather than the agent, who ought to shoulder the burden of
c ompensation f or these wrongs.
Call this the Responsibility A ccount of official immunity: state officials should not
ha v e to p a y c ompensation f or a wrongful harm they produc e when and be c ause the state (or
some state institution) bears the bulk of responsibility f or the production of that harm.
6
On
this ac c ount, the salient diff erenc e betw e en Firefighter damaging Neighbor’s home (Fire 1)
and Citizen damaging Neighbor’s home (Fire 2) is that Firefighter is not chiefly responsib le
f or the damage she does (it’s the state that is most responsib le) whereas Citizen is chiefly
responsib le f or the damage he does.
Some might obje ct to the idea that the b alanc e of responsibility c an c ome ap art from the
5
The w ord ‘responsibility’ is po lysemous, ha ving man y diff erent, closely-relate d meanings. There is one
sense of ‘responsibility’ on which it is tauto logous that the person who bears most of the responsibility f or a
harmful wrong is the person who has a duty to p a y c ompensation f or that wrong. Some p hilosop hers ha v e
c alle d this duty responsi bi lity . (Se e, f or e x amp le, Gary W atson, “Raz on Responsibility ,” Cri mi na l Law a nd
Phi losop hy 10 (2016): 295-409.) P a rt of what it is to be duty responsib le f or harm H is to ha v e a duty to pro vide
c ompensation f or H. The sort of responsibility I’m ref erring to in this p aper is diff erent. Call it agent respon-
si bi lity . Agent responsibility f or H c onsists in a f orm of c ontro l betw e en one’s agenc y and H; the tighter this
c onne ction, the greater one’s agent responsibility . Compensatory duty f or H is not p art of what it is to be agent
responsib le f or H, though the f act that some one has a c ompensatory duty f or H is often exp la i ned by the f act
that they are agent responsib le f or that harm. A similar distinction is made b y Josep h Raz, “Responsibility and
the Negligenc e Standard,” Oxford Jou rna l of Le g a l Stu di es , 30 (2010): 1-18.
6
A more e xtreme ac c ount in the neighborhood is what w e might c all the Vicarious Agency Account .
On this ac c ount, it isn’t just that the state bears the bulk of responsibility f or the harm in question, but that
it is the state (rather than the official) that is the agent of the harm; the official is merely a c onduit of state
agenc y . I w on’t spend an y time discussing this ac c ount, ho w ev er, as it has less prima f acie p lausibility than the
Responsibility Account and ho lds up ev en w orse against the obje ctions I press in this se ction.
117
b alanc e of c ausal c ontribution. But this clearly happens on oc c asion. Consider, f or e x amp le:
Overboa rd . Re ckless is driving his new spe e dbo at on a small lake. W anting to
se e ho w f ast his bo at c an go , he presses the throttle all the w a y f orw ard, bringing
his spe e d to more than twic e the legal spe e d limit. At such high spe e ds he f ails
to ev en notic e when his bo at c rashes into P addler’s c anoe, destro ying the c anoe
and thro wing P addler into the w ater. P addler c an sa v e herself from dro wning
only b y climbing onto a nearb y dock. But she c an only do so b y pulling herself
up b y the finger of Sle ep y , who is sunb athing at the end of the dock (and who
c annot be w o ken to giv e his c onsent). P addler pulls herself up , breaking Sle ep y’s
finger.
Sle ep y is o w e d c ompensation f or his bro ken finger. But although it is P addler who chose to
injure Sle ep y and who w as the dire ct c ause of S le ep y’s injuries, the burden of c ompensation
surely f alls on Re ckless. This, w e w ant to sa y , is be c ause Re ckless is the person who bears
chief responsibility f or Sle ep y’s injury .
The prob lem with the Responsibility Account isn’t that it makes a c onc eptual dis-
tinction betw e en c ausal c ontribution and responsibility . As the abo v e e x amp le sho ws, that’s
an important distinction. The prob lem is that the ac c ount undersells the e xtent to which
state officials are responsib le f or the harms they produc e. F or one thing, state officials are not
mere instruments of the state; they do not merely “run the c ode” written b y their superiors.
They often enjo y c onsiderab le disc retion as to how they f o llo w state dire ctiv es. Some harms
impose d b y state officials are non-discreti ona ry (in legal p arlanc e, “ministeria l ”): these are
harms that c annot f easib ly be a v oide d so long as the official f o llo ws the state’s dire ctiv es. But
man y harms are discreti ona ry : these are harms that are c ontingent on the official choosing one
118
w a y of f o llo wing state dire ctiv es rather than another. F or man y disc retionary harms , state of-
ficials e x ercise much more c ontro l o v er those outc omes than their princip als, and as such are
more responsib le f or producing those harms than their princip als. And y et the line betw e en
official immunity and official non-immunity is surely not the line betw e en disc retionary and
non-disc retionary harms: some disc retionary harms are among the p aradigm harms f or which
officials enjo y c ompensatory immunity . Re c all the c ase of Mistaken Imprisonment. Here,
w e note d, it se ems clear that Judge should not ha v e to c ompensate V ictim f or his wrongful
imprisonment. But suppose w e w ere to learn that Judge w as not legally re quire d to giv e an y
p articular sentenc e, and that it w as left to his disc retion to sentenc e V ictim to an ything be-
tw e en fiv e and ten y ears in prison. As it happene d, Judge sentenc e d V ictim to sev en y ears
in prison. Does this negate Judge’s immunity? Surely not. The intuitiv e v erdict that Judge
should not ha v e to c ompensate V ictim is not altere d b y learning that Judge e x ercise d disc re-
tion in the sentencing.
Here’s another reason w e should reje ct the Responsibility Account . Refle ct on our
attitudes to w ard the benefits produc e d b y state officials. Imagine that Firefighter is ordere d
on another mission. She is to ld to p arachute into a remote to wn to slo w the spread of an
appro aching f orest fire. She perf orms the mission e x actly as dire cte d and her eff orts sa v e the
liv es of dozens of pe op le. Although Firefighter sa v e d the to wnsf o lk be c ause she w as simp ly
f o llo wing orders, most of us w ould think it fitting to heap praise and honors on Firefighter.
W e w ould praise her, not just f or being good at her job , but espe cially f or the good outc ome
she brought about b y being good at her job; w e w ould praise her f or sa ving the liv es of the
to wnsf o lk. M ore o v er, w e w ould think that Firefighter deserv es greater praise f or sa ving those
pe op le than does the state itself. This is no doubt be c ause w e judge Jumper to be the agent
119
who is most responsib le f or the rescue. But if state officials bear chief responsibility f or the
benefits they produc e b y their obe dienc e, it is v ery hard to se e wh y they w ould not bear chief
responsibility f or the harms they impose b y their obe dienc e as w ell.
7
4.3 The Benefiary Pays Account
Neither the Promissory Ac c ount nor the Responsibility Ac c ount is p lausib le. As I c an se e
no other w a y in which the princip al-agent relation might e xp lain official immunity , I no w
turn to e xp lore elsewhere. In the ne xt tw o se ctions I e xp lore the idea that official immunity is
grounde d, not in a relation betw e en officials and the state, but instead in a relation betw e en
officials and the citi zen ry .
M an y p hilosop hers ha v e def ende d the so-c alle d Benefi cia ry P a ys Pri nci p le . There are man y
v ersions of this princip le, but the c ommon c ore is that
Beneficiary P a y s . The re c eipt of benefits resulting from wrongful harm c an
sometimes ground a duty on the p art of the beneficiary to pro vide c ompensa-
tion to t he victim of that harm.
8
7
Se e Jeff M cM ahan, Killing in W ar (Oxf ord: Oxf ord Univ ersity Press, 2009), 84-91.
8
Diff erent v ariants of this idea are def ende d b y Christian Barry and Robert Kirb y , “Sc epticism about Ben-
eficiary P a ys: A Critique,” Journal of App lie d Philosop h y 34 no . 3 (2017): 282-300; Sab a Bazargan-F orw ard,
“Grounding the Beneficiary P a ys Princip le,” Oxf ord Studies in P o litic al Philosop h y 8 (f orthc oming); Daniel
Butt, “‘ A Doctrine Quite New and Altogether Untenab le’: Def ending the Beneficiary P a ys Princip le,” Jour-
nal of App lie d Philosop h y 31 no . 4 (2014): 336-348; Ale x andra Cuoto , “The Beneficiary P a ys Princip le and
Strict Liability ,” Philosop hic al Studies 175 no . 9 (2018): 2169-2189; Robert Goodin, “Disgorging the Fruits of
Historic al W rongdoing,” Americ an P o litic al Scienc e Review 107 no .3 (2013): 478-491; Sigurd Lindstad, “Ben-
eficiary P a ys and Respe ct f or Autonom y ,” Social The ory and Practic e 47 no . 1 (2012): 153-169; T om P arr, “The
M o ral T ainte dness of Benefiting from Injustic e,” E thic al The ory and M oral Practic e 19 no . 4 (2016): 985-996;
T adros, “Orw ell ’s Battle,” 42-77; Judith Thomson, “Pref erential Hiring,” Philosop h y and Pub lic Aff airs 2 no .
4 (1973): 364-384; and Daniel V iehoff, “Legitimac y as a Right to Err,” in NOMOS LXI: P o litic al Legitimac y ,
e ds. Jack Knight and M elissa Sch w artzberg (New Y ork: NYU Press, 2019), 174-200.
120
Def enders of Beneficiary P a ys typic ally claim that the beneficiary’s agenc y ne e dn’t bear an y
c onne ction to the wrong or to their o wn benefit. The beneficiary ne e d not ha v e c ause d or
c ontribute d to the wrong. The beneficiary ne e d not ha v e aske d f or or intende d the benefit.
No authorization is ne e de d. The beneficiary ne e d only rec eive the benefit.
It’s natural to think to e xp lain official immunity in terms of something like Beneficiary
P a ys. As f ormulate d, ho w ev er, the princip le p lainly w on’t do the job . The princip le f ocuses
on the beneficiaries of pa rti cu la r instanc es of wrongful harm. But there are man y instanc es
where state officials c ause wrongful harm without benefiting an y one — instanc es where the
official se ems nonetheless immune from o wing c ompensation. No one benefits from the ac-
tions of Judge or Guard in Mistaken Imprisonment, f or instanc e. And w e c an easily imagine
a v ariant o f Fire 1 in whic h no o ne benefits from Firefighter’s action: perhaps she isn’t ab le to
e xtinguish the fire despite all the w ater she b lasts on the house. The absenc e of benefit in these
instanc es doesn’t matter: whether Judge, Guard, or Firefighter should ha v e to p a y c ompen-
sation f or the harms they impose is not c ontingent on whether those p articular harms benefit
an y one.
The upshot is that if w e are going to appeal to the benefits of the actions of state officials to
e xp lain official immunity , w e ne e d to appeal, not to the benefits that result from pa rti cu la r
harms, but rather to the benefits that result from some broader project that giv es rise to the
harms in question. The idea is this. When some one produc es a wrongful harm in the e x e cu-
tion of a proje ct from which the Xs benefit, the Xs thereb y incur duties to bear (at least some
of) the burden of c ompensating f or that harm. The state’s proje ct of maintaining a just and
good po litic al order is a proje ct from which its citizens generally benefit. This — the story
goes — is wh y the citizenry is duty-bound to shoulder the c osts of c ertain official wrongs.
121
Call this the Beneficial St a te A ccount .
9
On this ac c ount, the c rucial diff erenc e be-
tw e en Fire 1 and Fire 2 is the f act that Firefighter’s actions (in Fire 1) are p art of a bro ader
proje ct that serv es the citizenry , whereas Citizen’s actions (in Fire 2) are not.
Let’s interrogate the idea at the heart of this ac c ount — the idea that duties of c ompensa-
tion c an f all on the beneficiary of a wrongful harm simp ly beca use they are a beneficiary . This
idea is clearly imp lausib le, absent some kind of restriction. T o se e wh y , c onsider, f or e x amp le:
Three Fis hermen . There are thre e fishermen (Al, Bob , and Chuck) that share
the shoreline of a b a y and c ompete f or business. In the dead of night and in a
drunken fit of jealous y , Al sneaks onto Chuck’s dock and burns Chuck’s bo at.
While making his esc ape, ho w ev er, Al slips and breaks his arm. As a c onse-
quenc e of these ev ents, neither Al nor Chuck are ab le to fish f or man y w e eks.
This helps Bob to c atch more fish and makes a greater profit than he otherwise
w ould ha v e.
T w o things are ob vious. The first is that Al has a duty to c ompensate Chuck f or destro ying
his bo at. The se c ond is that Bob — the only beneficiary of Chuck’s loss — has no duty to
assume Al ’s c ompensatory burden. Neither Al nor Chuck c an legitimately demand this of
Bob .
Compensatory duty clearly re quires more than the mere re c eipt of benefit. If the “benefi-
ciary p a ys ” idea is to be at all p lausib le, w e ne e d some w a y of restricting the idea to a c ertain
subset of beneficiaries. Can w e loc ate a p lausib le restriction?
9
It should be note d that man y v ersions of the Beneficiary P a ys Princip le will not get us an ything like this
ac c ount. F or e x amp le, some the orists (e.g., Bazargan-F orw ard, Goodin, and P arr) def end only a disgorgement
v ersion of the princip le, ac c ording to which beneficiaries of a harm ne e dn’t c ontribute to offsetting the victim’s
losses, but ne e d only transf er their o wn g a i ns to the victim.
122
One possibile restriction takes its inspiration from a f amous quote from Ra wls regarding
duties of “f air p la y”:
When a number of persons engage in a mutually adv antage ous c ooperativ e v en-
ture ac c ording to rules, and thus restrict their liberty in w a ys ne c essary to yield
adv antages f or all, those who ha v e submitte d to these restrictions ha v e a right
to similar ac quiesc enc e on the p art of those who ha v e benefite d from their sub-
mission.
10
Ra wls is talking about duties that beneficiaries ha v e to bear the c osts of o bedi enc e , not the
c osts of c ompensation. But w e might think that his w a y of restricting this princip le is also a
nic e w a y of restricting the “beneficiary p a ys ” idea. P erhaps there is something spe cial about
c ooper ative proje cts; perhaps one does not ac quire duties of assumption b y benefiting from
just an y wrong, but only from wrongs that are the products of c ooperativ e proje cts.
The c ooperativ e-proje ct restriction doesn’t help , ho w ev er. There are all sorts of c oopera-
tiv e proje cts f or which beneficiaries p lainly do not ha v e duties of c ompensatory assumption.
Consider, f or e x amp le:
Be l ls . A small minority of pe op le in y our neighborhood ha v e installe d large
church bells on their properties, and ha v e be en w orking together o v er the p ast
f ew months to produc e beautiful music at the top of ev ery hour, to the enjo y-
ment of ev ery one in the neighborhood. One night, ho w ev er, while one of the
bell-ringers is w orking the bells, the rope fra ys. The bell is about to f all on Bell
10
Jo hn Ra wls, A Theory of Justi c e (Cambridge, M assachusetts: Harv ard Univ ersity Press, 1971), 108-114.
W e get a v ery similar princip le from H.L.A. Hart, “ Are Th ere An y Natural Rights?” Phi losop hi ca l Revi ew 64
no . 2 (1955): 185.
123
Ringer. He a v oids being c rushe d in the only w a y he c an — b y re dire cting the
bell onto hi s neighbor’s unoc cupie d c ar.
11
Bell Ringer clearly incurs a duty to c ompensate his neighbor f or the damage to her c ar.
Does ev ery one in the neighborhood who has benefite d from the to lling of the bells o w e it to
Bell Ringer to offset his duty? Intuitiv ely , no . (In the ne xt se ction I’ll e xp lore the relev ant
diff erenc e betw e en state officials and the bell ringers.)
Here is a diff erent restriction w e might try . W e might dep art from the standard emp ha-
sis on benefits that are i nvo lu nta ri ly re c eiv e d, and insist that beneficiaries take on duties of
assumption only when they ha v e in some sense ac c epted the benefits.
12
But this mo v e w on’t w ork either. In Thre e Fishermen, f or e x amp le, Al ’s wrong makes
it so that Bob has the opportunity to c atch more fish. But Bob re c eiv es the benefits of this
opportunity only b y going out on the w ater and intentionally hauling in the e xtra fish. The
benefit is not droppe d in his lap; he goes out and claims it. What’s more, this benefit is such
that Bob c ould easily a v oid it. W e c an imagine that it w ould be no troub le to a v oid c atching
e xtra fish. And y et despite all this Bob does not ha v e a duty to offset Al ’s c ompensatory
burden.
If an ything, the sense in which citizens “ac c ept” the benefits of the state is less robust than
the sense in which Bob ac c epts his benefits. It is eas y f or Bob to a v oid c atching e xtra fish.
It is not eas y to a v oid the benefits that are c onf erre d on citizens b y the state. M ost of these
benefits c an be a v oide d only at tremendous c ost — f or e x amp le, emigration, social iso lation,
or po v erty .
11
W ith a bo w to Robert N ozick’s c ase of the neighborhood P A s ystem from A na rc hy, State, a nd Utopia
(W iley-Blackw ell 2001): 93-95.
12
Se e A. Jo hn Simmons, “The Princip le of F air Pla y ,” Phi losop hy a nd Pu b li c Aff a i rs 8 no . 4 (1979):307-337
f or an attempt to f ormulate the princip le of f air p la y in terms of ac c epte d benefits.
124
Let’s take one more stab at restricting the “beneficiary p a ys ” idea. Instead of f ocusing only
on f eatures of the re cipient (such as what benefits they ha v e re c eiv e d and ho w they ha v e re-
c eiv e d those benefits) w e might also re quire that c ertain c onditions be met b y the agent of the
wrong. Daniel V iehoff has re c ently def ende d a v ersion of this idea.
13
He claims that the ben-
eficiaries of wrongful harm ac quire c ompensatory duties only when the agent who imposes
the harm satisfies c ertain “ deliberativ e” c onditions:
Where [the agent] del iberates in the right w a y , [the agent] acts f or [the benefi-
ciary], not just in the sense of benefiting [her], but in the sense of adopting [her]
practic al position. [The agent] adopts [the beneficiary’s] practic al position b y
acting only on the reasons that [she] has — [her] aims and proje cts, a nd of t he
rig hts of ot hers i nsof a r as t hey li mit t hei r pu rsuit ... In practic e, this means that
[the agent] must, f or instanc e, choose what c ourse of action to take b y c onsid-
ering only ho w diff erent c ourses aff e ct [the beneficiary’s] proje cts and aims, not
b y ho w the y aff e ct [the agent’s] o wn.
14
V iehoff’s propose d restriction does help us to a v oid the imp lausib le result that Bob o w es
c ompensation in Thre e Fishermen. This is so ev en if w e suppose that Al burns Chuck’s
bo at, not out of jealous y , but so lely with the go al of benefitting Bob . Al w ould still f ail to
me et V iehoff’s deliberativ e c ondition, sinc e Al w ould f ail to take ac c ount of the w a y in which
Chu c k’s rig hts limit the w a ys in which Bo b ’s i nterests c an be pursue d (se e the italicize d portion
of the V iehoff quote abo v e).
13
Daniel V iehoff, “Legitimac y as a Right to Err,” in NOMOS LXI: Po liti ca l Legiti macy , e ds. Jack Knight
and M elissa Sch w artzberg (New Y ork: NYU Press, 2019), 174-200.
14
“Legitimac y ,” 192.
125
V iehoff’s proposal deliv ers this upside, ho w ev er, at the c ost of greatly re ducing the e x-
p lanatory reach of the Beneficial State Account . Consider, f or e x amp le, a v ariant of
F einberg’s w ell-kno wn Cabin c ase:
15
Fi re Ca bi n . Firefighter is a “smo kejumper” with the F orest Servic e, who has
p arachute d onto a mountain to prev ent the spread of a dangerous f orest fire.
She is injure d in the drop , sep arate d from her team, and c an surviv e only b y
breaking do wn the door of Owner’s unoc cupie d c abin to find me dic al supp lies
to b andage her w ounds. Jumper breaks do wn the door.
Although Firefighter acts without f ault in this c ase, c ompensation is p lainly o w e d f or the
damages to the c abin. And y et, just as it is clear that Firefighter should not ha v e to p a y c om-
pensation f or the damages she c auses Neighbor in Fire 1, it is likewise clear that Firefighter
should not ha v e to p a y c ompensation f or the damages she c auses Owner in this c ase. V iehoff’s
deliberativ e restriction on the Beneficial State Account , ho w ev er, does not deliv er this
v erdict. Firefighter does not break into the c abin f or the sake of an y one but herself. The rea-
sons that motiv ate — and justify — her tresp ass are f acts about her o wn w ell-being. V iehoff’s
ac c ount, then, doesn’t e xp lain wh y official immunity e xtends to c ases such as this.
No w w e might try relaxing V iehoff’s propose d restriction so that Jumper ne e d only adopt
some one else’s practic al position with respe ct to the missi on t h at gives rise to her trespass .
P erhaps it is enough that she acts only f or other-regarding reasons when she p arachutes onto
the mountain, and it is this f act that renders her intende d beneficiaries liab le f or damages that
result from the mission — ev en if those damages are more dire ctly relate d to Jumper’s acting
15
Joel F einberg, “V o luntary Euthanasia and the Inalienab le Right to Lif e,” Phi losop hy a nd Pu b li c Aff a i rs 7
(1978): 93-123.
126
on the b asis of her o wn w ell-being.
W e might be ab le to c apture the intuitiv e v erdict in Fire Cabin b y such a mo v e. But troub le
remains. F or one thing, this relax e d restriction will not help us a v oid some of the prob lems
w e’v e already se en f or the “beneficiary p a ys ” idea. Consider, f or e x amp le, a v ariant of Bells
in which the reasons f or which Bell Ringer has p articip ate d in the bell proje ct are entirely
other-regarding. Unbekno wnst to his ne ighbors, he has de dic ate d his lif e to pro viding others
with beautiful and delightful music. By the relax e d restriction under c onsideration, then, the
damage he c auses to his neighbor’s c ar should be analogous to the harm Firefighter c auses to
the o wner of the c abin: the beneficiaries of Bell Ringer’s “mission” should assume this c osts
of that mission. This is v ery c ounterintuitiv e.
But ev en if w e are willing to ac c ept such c ounterintuitiv e results, the V iehoff proposal just
does not e xtend w ell to an e xp lanation of official immunity . This is the most serious prob lem
f or the proposal: it dra ws the line betw e en official immunity and official non-immunity in
an imp lausib le p lac e. Imagine, f or e x amp le, a Kantian who is motiv ate d to serv e as a judge
purely b y the idea that judicial servic e is the best w a y f or him to do his impersonal moral duty .
Or imagine a firefighter who has purely religious motiv ations f or her servic e; she believ es she
c an best c omp ly with the c ommands or desires of God b y prote cting pe op le from f orest fires.
Neither of these officials take up the practic al position of the citizenry . And thus V iehoff’s
proposal tells us — v ery imp lausib ly — that such officials do not enjo y immunity f or an y of
the damages they impose in the line of duty , no matter ho w imp artially they app ly the la w
or ho w e x c ellently they fight fires. W e should ac c ept such a seriously revisionary ac c ount of
official immunity only if a less-revisionary ac c ount is una v ailab le.
127
4.4 The Price of Duty Account
While I lack the sp ac e to c onsider ev ery c onc eiv ab le w a y of restricting the Beneficial State
Account , I hope to ha v e sho wn that the most natural c andidates f ail. The Beneficial
State Account does not loo k p lausib le. But our refle ction on the ac c ount does point the
w a y to w ards a better, less-revisionary ac c ount of official immunity .
The Beneficial State Account has prima f acie appeal, I think, be c ause it is a natural
w a y of pre cisifying the intuitiv e idea that official immunity is grounde d in c onsiderations
of distributiv e f airness. It gets its appeal from the attractiv e idea that citizens ha v e duties
to assume some official liability be c ause it w ould be u nf a i r if officials had to bear all those
burdens on their o wn. In this se ction I propose an alternativ e w a y of unp acking this attractiv e
idea.
I w ant to start b y returning to the earlier quote from Ra wls. In that quote Ra wls proposes
a restricte d v ersion of Beneficiary P a ys. He doesn’t claim that mere benefit puts one under a
duty to assume c ertain c osts; he claims that the benefits of c ooper ative projects put one under a
duty of assumption. I argue d abo v e that this isn’t quite right. There are c ooperativ e proje cts
— like the to lling of the bells — f or which beneficiaries are not re quire d to offset c osts that
f all on the p articip ants. Cost sharing is not re quire d of just an y beneficial c ooperativ e proje ct.
That said, I think Ra wls ’s f ocus on c ooperativ e proje cts puts us on the right track. There
is something spe cial about a c ertain class of c ooperativ e proje cts. Consider:
Grou p Rescu e . A hiker’s lif e is threatene d b y an une xpe cte d sno wstorm. Alic e
and a group of other strangers stumb le ac ross him. They c an rescue him only
b y w orking together to c arry him into Owner’s c abin. Alic e runs ahead to open
128
the c abin, only to find the door and windo ws locke d. She breaks do wn the
door, and the group arriv es shortly thereafter to c arry the hiker into the c abin.
Unf ortunately , the hiker’s health does not impro v e and he p asses a w a y in the
night.
Comp are this c ase with Bells. W e note d earlier that Bell Ringer is not o w e d indemnific a-
tion b y the beneficiaries of the bells. But it also se ems that he is not o w e d an y indemnific ation
b y the other p articip ants of that proje ct, absent prior agre ement to share such c osts. Things
are p lainly diff erent in the c ase of Group Rescue. The c abin o wner is o w e d c ompensation
f or his bro ken door. But f or reasons that are not imme diately app arent, it se ems clear that
Alic e should not ha v e to bear that burden alone. The other members of the rescue team must
share this burden with her, regardless of whether there w as prior agre ement to do so .
16
This is a surprising diff erenc e. The f act that Alic e, but not Bell Ringer, enjo ys a right
to indemnific ation has nothing to do with the f act that she imposes the harm permissi b ly .
Bell Ringer also acts permissib ly . Both enjo y a lesser-evil justific ation f or their action — Bell
Ringer to sa v e his o wn lif e, Alic e to sa v e some one else’s. Nor c an w e e xp lain Alic e’s right
to indemnific ation b y appeal to the rec ei pt of benefits . Alic e’s burden must be offset b y her
c o-rescuers, but her c o-rescuers benefit neither from Alic e’s act ions nor from the rescue as a
who le.
P erhaps it might se em that the key diff erenc e is a sort of c omp li city . Although Alic e is the
person most responsib le f or the door’s destruction, her ac c omp lic es are (permissib ly) c om-
p licit in this tresp ass b y w a y of their c ontributing to other aspe cts of the proje ct of which this
tresp ass is a p art. It might se em that they must share in the c ompensatory burden f or the same
16
At least this c ost should be share d on the assumption that other c osts associate d with the rescue ha v e be en
distribute d f airly .
129
reason that the geta w a y driv er at a b ank robbery might be re quire d to share in c ompensating
those injure d in the robbery .
17
While it is p lausib le that c omp licity is one w a y b y which persons c an incur duties of indem-
nific ation, I don’t think this c an e xp lain wh y Alic e should not ha v e to bear the c ompensatory
burden alone. F or one thing, it’s not clear wh y Alic e’s c o-rescuers w ould be c omp licit in the
relev ant sense in Group Rescue, but Bell Ringer’s c o-ringers not likewise c omp licit in Bells.
It’s not clear that c omp licity gets us the diff erenc e w e ne e d betw e en these tw o c ases. Here’s a
se c ond prob lem f or the c omp licity proposal. Consider a v ariant of Group Rescue in which
one member of the group that stumb les upon the hiker de cides not to help . Call him Egoist.
Egoist’s help , let’s suppose, is important to the rescue eff ort, and he c ould pro vide it at mini-
mal c ost to himself. He is thus under a duty to p articip ate in the rescue. But he w ould pref er
to get home to hot c oc o a and a w arm fire, and so he lea v es the group behind.
Egoist is clearly not c omp licit in Alic e’s tresp ass, sinc e he doesn’t p articip ate in the rescue.
But just as Egoist has a duty to p articip ate in the rescue in the first p lac e, he likewise se ems
to ha v e a duty to share in the c osts of the rescue — regardless of whether he discharges his
ante c e dent duty of p articip ation. M ore o v er, it se ems that his duty of p articip ation is the only
thing that c an e xp lain his duty of c ost sharing: Egoist is morally re quire d to share in the c osts
of the rescue pre cisely beca use he w as morally re quire d to p articip ate in that rescue in the first
p lac e.
The operativ e princip le here is what I will c all
The Pric e of Duty . Ceteris peribus, when the Xs are morally re quire d to p ar-
ticip ate in a c ooperativ e proje ct, the Xs are re quire d to share in c ertain c osts of
17
F or a discussion of the c onc ept of c omp li citous lia bi lity se e Sab a Bazargan-F orw ard, “Comp licitous Liabil-
ity in W ar,” Phi losop hi ca l Stu di es 165 (2013): 177-195.
130
that proje ct (regardless of whether they actually p articip ate).
18
This princip le of distributiv e justic e is not limite d to the c ompensatory c osts of morally-
ob ligatory proje cts. It e xtends to all sorts of proje ct c osts. Suppose, f or e x amp le, that y ou
and I belong to a small village in a remote desert. A drought has f allen on the region, and all
the members of our village will die if they do not get fresh w ater, and soon. They c an get fresh
w ater only if I use m y unique c onstruction kno wle dge to build a small dam and y ou use y our
unique chemistry kno wle dge to build a w ater-purific ation devic e. I c an build the dam with
either mud bricks or c onc rete b locks. Building with the mud bricks will c ost me fiv e da ys of
labor; building with the hea vier c onc rete b locks will c ost me ten da ys of labor. If I use the
mud bricks, ho w ev er, it will make the w ater much dirtier, and will re quire greater w ork on
y our p art. If I use mud bricks, it will take y ou fifte en da ys to build y our w ater-purific ation
machine; if I use c onc rete b locks, it will take y ou ten da ys to build y our w ater-purific ation
machine.
Let’s assume that y ou are duty-bound to bui ld the w ater-purific ation devic e, that I am
duty-bound to build the dam, and that our duties are e qually demanding. Grante d this as-
sumption, I find it intuitiv ely c ompelling that I owe it to you to bui ld wit h c oncrete r at her t ha n
mu d . If I build the dam with mud, I w ould wrong y ou in that I w ould manage m y duty to
c ontribute to the proje ct of pro viding fresh w ater f or our village in a w a y that makes it more
c ostly (both c omp arativ ely and non-c omp arativ ely) f or y ou to discharge y our same duty .
In the same w a y , I claim, the other p articip ants in Group Rescue w ould wrong Alic e if they
f aile d to share in her c ompensatory burden. By f ailing to share in that burden, they w ould
18
Suppose the reason Hiker w as in ne e d of rescue w as that V illain had injure d Hiker b y intentionally pushing
him off a cliff. Giv en V illain’s culp ability f or the ne e d f or rescue in the first p lac e, it is V illain who should bear
the c osts of the rescue, and not the rescuers. Cases like these are wh y Pric e of Duty is a ‘ c eteris peribus ’ princip le.
M o re on the qualifie d status of this princip le, and more on the signific anc e of the ‘ c ertain’ qualifier, in §4.5.
131
manage their duties to c ontribute to the rescue of Hiker in a w a y that made it much more
c ostly f or Alic e to discharge her same duty .
The Pric e of Duty princip le nic ely ac c ommodates our judgments about ho w the c osts of
morally ob ligatory proje cts should be distribute d. It also makes sense of the intuitiv e diff er-
enc e betw e en c ases like Group Rescue and Bells. In Group Rescue, the damages c ause d b y
Alic e are the f allout of a proje ct that other pe op le are re quire d to p articip ate in. In Bells, b y
c ontrast, the damages c ause d b y Alic e are the f allout of a proje ct that is morally opti ona l f or its
p articip ants. There’s no moral duty to install and ring bells f or y our neighborhood to enjo y .
Pric e of Duty is a c ompelling princip le of distributiv e justic e. And it is a princip le that
e xp lains official immunity . This is be c ause the po litic al proje ct is more like the proje ct of res-
cuing the strande d hiker than like the proje ct of ringing the bells: the citizens of (reasonab ly
just) states are morally re quire d to p articip ate in, and support, the po litic al proje ct. By Pric e
of Duty , citizens of a state are re quire d to share in c ertain c osts of that proje ct. This entails
a c ertain amount of official immunity , sinc e state officials w ould bear more than their f air
share of the c osts of the po litic al proje ct if they w ere made to bear the full brunt of all the
c ompensatory c osts they incur. Not to be c onfuse d with the Pric e of Duty princip le itself,
let’s c all this e xp lanation of official immunity the Price of Duty A ccount .
Notic e that although this ac c ount does not re quire that a giv en citizen benefit from an y
p articular act of tresp ass, benefit still has some ro le into the s tory . F or one thing, e xpe cte d
benefits c an aff e ct whether a proje ct is morally re quire d. The rescuers in Group Rescue,
f or e x amp le, are morally re quire d to p articip ate in the rescue of hiker only be c ause of this
proje ct’s prospe cts f or sa ving the hiker’s lif e. Had the e xpe cte d benefits be en much less, the
proje ct w ould ha v e be en morally optional. Likewise, citizens are only re quire d to p articip ate
132
in the po litic al proje ct when and be c ause that proje ct promises to prote ct c ertain rights and
c onf er c ertain benefits (e.g., prote ction of body and property , a s ystem of c orre ctiv e justic e,
ac c ess to essential servic es). Benefits matter in other w a ys as w ell. F or e x amp le, if some one
benefits more from the po litic al proje ct than others (or less), this ma y aff e ct what share of
the c osts of that proje ct they must bear. If Alic e w ere the only member of the rescue team in
Group Rescue to re c eiv e a c ash rew ard f or the rescue, this w ould p lainly alter the share of the
c ompensatory c osts she should bear.
The f act that state officials benefit the citizenry is thus not irrelev ant to official immunity .
But benefit p la ys an i ndi rect ro le. What fundamentally matters is the f act that the po litic al
proje ct is one that citizens are morally re quire d to p articip ate in. Official immunity is just
what it loo ks like to f airly distribute the c osts of this mandatory proje ct.
4.5 Limits
W e ha v e some pre-the oretic intuitions about the limits of official immunity . I’v e relie d on
some of these intuitions throughout this p aper. F or e x amp le, I’v e appeale d to the widely
held belief that a f air and c onscientious judge who makes a reasonab le mistake in sentencing
an innoc ent man to prison should not shoulder the c osts of c ompensating the victim. This is
a clear c ase of official immunity . And of c ourse there are clear c ases of official non-immunity .
If a po lic e offic er takes a b aseb all b at to the legs of his romantic riv al, he is p lainly not immune
from o wing his victim c ompensation, regardless of whether this tresp ass oc curre d while he
w as “ on duty”.
But betw e en the clear c ases of official immunity and official non-immunity is a large penum-
bra. Dire ct refle ction on c ases of official tresp ass c an only get us so f ar in disc erning the li mits
133
of official immunity . A c entral c ontention of this p aper is that w e c an get further b y better
understanding the grou nds of official immunity . I’v e thus f ar argue d that official immunity is
the product of the F air Pric e of Duty princip le, which claims that persons who are morally re-
quire d to p articip ate in a c ooperativ e proje ct are re quire d to share c ertain c osts of that proje ct.
Be c ause this princip le is a v ery general one, w e c an make progress in thinking about the limits
of official immunity b y refle cting on much simp ler e x amp les of morally re quire d proje cts.
Our intuitions are more clear in these c ases, and easier to interpret.
When w e refle ct on such c ases, f or what sorts of wrongful harm do w e find that duties of
c ost sharing app ly? One thing w e find is that duti es of c ost s ha ri ng a re not li mited to i nsta nc es
of permissi b le trespass . Consider a v ariant of Group Rescue: as the group drags Hiker inside
the c abin, Alic e misperc eiv es the distanc e betw e en her body and an e xpensiv e glass v ase on a
nearb y tab le, knocking it o v er. Destro ying the v ase is not permissib le in a f act-relativ e, belief-
relativ e, or evidenc e-relativ e sense.
19
And y et, pro vide d w e imagine that Alic e’s misperc eption
is not the result of an y negligenc e or lack of c are, it se ems that Alic e should not ha v e to bear
this c ompensatory burden alone. (These c ases also sho w that duties of c ost sharing e xtend to
some harms that do not c ontri bute to the ob ligatory proje ct’s end.)
There are f our sorts of harm, ho w ev er, f or which duties of c ost sharing do not se em to
e xtend. First, there is what I will c all i nci denta l harm. Suppose, f or e x amp le, that a member
of the rescue team de cides to enjo y a cigarette after helping pull the hiker inside the c abin. The
cigarette slips from his hand, f alls betw e en the c racks of the porch, and ignites the drie d lea v es
beneath, c ausing e xtensiv e damage to the porch. These damages se em importantly diff erent
from the damages Alic e c auses to the door. Intuitiv ely , the smo ker does not ha v e a claim that
19
The distinction betw e en these thre e kinds of permissibility is c ommonly attribute d to Derek P arfit, On
What Matters , v o l. 1 (Oxf ord, 2011): 150.
134
his c o-rescuers help shoulder the c osts of the damage to the porch f or the reason that these
c osts c annot p lausib ly be desc ribe d as c osts “ of” the rescue proje ct. It is not reasonab le f or
the smo ker to believ e that a cigarette break w ould c ontribute in an y w a y to the rescue proje ct,
nor does his p articip ation in the rescue proje ct make it difficult or c ostly to a v oid smo king on
this oc c asion.
20
The mere f act that damages are impose d whi le some one is p articip ating in an
ob ligatory proje ct does not suffic e to attribute those damages to the proje ct.
This suggests that official immunity does not e xtend to harms that are “incidental ” to the
po litic al proje ct. Citizens do not o w e an ything to po lic e offic ers who damage priv ate property
while driving to the donut shop . The f act that a po lic e offic er is “ on the clock” does not suffic e
to render him immune from such damages; citizens are not re quire d to share in the c osts of
an official ’s harm when the harm is unrelate d to the official ’s mission and the official ’s mission
does not put him in a position that makes it v ery c ostly to a v oid such harm.
A se c ond type of harm to which duties of c ost sharing do not app ly is neg ligent harm.
Suppose that in Group Rescue, Al ic e c ould ha v e simp ly turne d the door handle rather than
smash do wn the door, and that the only reason she didn’t try turning the door handle w as that
she w as ap athetic to the c abin o wner’s interests, or that she let her o wn lo v e of door kicking
sw amp her c onc ern f or the c osts this w ould impose on others, or that she w as just being lazy .
When w e tell the story in this w a y , Alic e clearly lacks a claim against her c o-rescuers to their
c ompensatory help . These c osts are “ on her”; they are c osts that w ere not merely a v oidab le,
20
Intuitiv ely , the c osts of the cigarette fire are the smo ker’s to bear alone, whereas the c osts of Alic e’s ac ci-
dental breaking of the v ase should be share d. What e xp lains this diff erenc e? T w o things, I think. First, whereas
the choic es and actions that most dire ctly lead to the cigarette fire are not p lausib ly thought of c onstitu ents of
the rescue proje ct, the choic es and actions that most dire ctly lead to the bro ken v ase are. Se c ond, whereas the
smo ker’s p articip ation in the rescue proje ct does not put him in a position where he c annot easily a v oid c reating
a fire, Alic e’s p articip ation in the rescue proje ct does put her in a position where she c annot easily a v oid breaking
the v ase.
135
but that w ould not e xist had Alic e treate d others with the c are she o w e d them. The members
of an ob ligatory proje ct do not ha v e a duty to share in the c osts that result from another
member’s f ailure to act with due c are. This makes sense: y ou c annot reasonab ly demand
that I not make an y mistakes in p articip ating in our share d proje ct, but y ou c an reasonab ly
demand that I e x ercise due c are to a v oid mistakes.
21
This suggests that official immunity does not e xtend to negli gent harm. Citizens, f or e x-
amp le, do not o w e an ything to the judge who sentenc es an innoc ent person to prison be c ause,
in his haste to get home f or dinner, he loo ks only at evidenc e presente d b y the prose cution
and ignores evidenc e presente d b y the def ense.
A third t ype of harm f or which duties of c ost sharing se em not to app ly is what I will c all
reneg ade harm. Suppose Alic e and her c o-rescuers agre e on a p lan as to ho w they will break
into the c abin. They all agre e on a p lan to break a windo w rather than the e xpensiv e door, and
they send Alic e ahead to do the de e d. When Alic e arriv es at the c abin, ho w ev er, she de cides to
deviate from the p lan, on her o wn belief that it will be easier f or the group to use the door than
the windo w . She kicks in the door. This might not be an e x amp le of negligenc e on her p art:
w e might suppose that she breaks do wn the door pre cisely be c ause she thinks this will be in
ev ery one’s best interest, and w e might suppose that she is right about this. Likewise, Alic e’s
action is not disassociate d from the proje ct. Her action is c onne cte d up with the proje ct in
the most straightf orw ard w a y: it c ontributes to the proje ct’s go al. And y et it se ems that Alic e
c annot demand that her c o-rescuers help her bear the e xtra c osts that she imposes b y breaking
the door rather than the windo w . This is pre cisely be c ause she deviate d from the p lan that
21
There are c omp lic ations that arise from the f act that negligenc e c omes in degre es. Does the smallest amount
of negligenc e negate all immunity? Or does the degre e of c ompensatory help an agent is o w e d lessen as the degre e
of negligenc e inc reases? Or is there some thresho ld of negligenc e abo v e which duties of c ost sharing disappear?
136
w as agre e d upon b y the group .
22
This suggests that official immunity does not e xtend to renegade harm. I lack the sp ac e
to dev elop a full ac c ount of what c onstitutes renegade c onduct b y a state official, but the
f o llo wing se ems p lausib le: in a state that has settle d legal rules that are the products of demo-
c ratic proc e dures, these rules c onstitute p art of the “p lan” that state officials must subje ct
themselv es to .
23
F or e x amp le, in the Unite d States it is p art of the national p lan that f e deral
agents ma y in v ade a person’s home only if they ha v e a w arrant. If a f e deral agent in v ades y our
neighbor’s home without a w arrant, and his in v asion re quires him to damage y our door, it
doesn’t matter ho w c areful and efficient the agent acte d in the c ourse of the in v asion. He
o w es c ompensation f or the damage to y our home, and he is not o w e d indemnific ation f or
those c osts. The f e deral agent is not o w e d indemnific ation be c ause his tresp ass is f orbidden
b y the agre e d-upon p lan f or ho w the U.S. po litic al proje ct is to be pursue d. The c osts of
renegade activity f all outside the sc ope of proje ct c osts that must be share d.
I’ll c onclude this se ction b y c onsidering an espe cially challenging c ase: the enf orc ement of
unjust la ws. These are not c ases in which an official deviates from the p lan. These are c ases
in which an official f o llo ws the p lan, but where the p lan itself is mistaken. Must the citizenry
share in the c osts of the enf orc ement of unjust la ws?
Refle ction on simp le c ases of group rescue suggests that some, but not all, of the c osts of
the enf orc ement of unjust la ws should be share d b y the citizenry . Suppose nine of us are
p lanning ho w to rescue the hiker. Thre e p lans are propose d. F our members propose that
22
It is too strong to sa y that duties of c ost sharing app ly only to tresp asses that are strictly “p art of the p lan”.
These duties also app ly to damages that result from reasona b le ada ptati ons of a p lan. F or e x amp le, Alic e w ould
be o w e d c ompensatory help if she arriv e d at the c abin to find the windo w impenetrab le, and f or this reason
kicke d in the door instead.
23
Se e Sc ott Shapiro , Leg a lity (Harv ard, M assachusetts: Harv ard Univ ersity Press, 2011) f or a view ac c ording
to which a s ystem of la w just is a sort of p lan.
137
w e break into the unoc cupie d c abin. F our other members propose that w e attempt to bring
hiker do wn the mountain on a sle d. Neither of these tw o options is ob viously better than the
other, in light of the c ommon inf ormation a v ailab le to the group . The ninth member of the
group proposes a surprising third p lan. He proposes that w e thro w the hiker off a cliff, in his
sinc ere belief that angels will c atch the hiker and fly him to saf ety .
Suppose that the ninth member of the group — a true master of the rhetoric al arts —
someho w manages to c on vinc e a ma jority of the group to go along with his p lan. Eight mem-
bers of the group proc e e d to thro w the hiker off the cliff, with only one person refusing to
p articip ate. As it turns out, the hiker is not c aught b y angels, but instead f alls to his death,
his body c rashing into and injuring a b ystander at the bottom of the cliff. This much se ems
clear: c ompensation is o w e d f or the b ystander’s injuries, but the one person who did not
p articip ate in the killing is not re quire d to share in this c ompensatory burden.
But no w suppose instead that the group de cides on the p lan to break into the c abin. Eight
members of the group go ahead and break into the c abin, which re quires breaking do wn the
c abin door. One person abstains from p articip ating. As it turns out, ho w ev er, the ma jor-
ity imp lements a suboptimal p lan: had they chosen to take hiker do wn the mountain on a
sle d, they c ould ha v e rescue d the hiker without an y tresp ass whatsoev er. Does this mean, as
in the previous iteration of the c ase, that the lone non-p articip ant ne e dn’t share in the c om-
pensatory burden? Here things se em v ery diff erent. It se ems the dissenter s hou ld share in
the c ompensatory burden, pre cisely be c ause this burden is the result of a reasonab le mistake.
Giv en the inf ormation pub licly a v ailab le to the group , it w as reasonab le f or them to disagre e
as to whether the c abin-in v asion p lan or the sle dding p lan w as best. By c ontrast, it w as not
reasonab le f or them to disagre e as to whether it w as best to toss the hiker off the cliff.
138
This is suggestiv e of a w a y to think about official immunity in the c onte xt of the enf orc e-
ment of unjust la ws. Official immunity e xtends to the c osts of the enf orc ement of reasona b le ,
unjust la ws. Some unjust la ws are “reasonab le” in the sense that, although they are in f act un-
just, it is reasonab le f or pe op le to disagre e about whether they are just, relativ e to some body
of pub lic inf ormation.
24
Citizens ha v e a duty to share in the c ompensatory c osts of the (non-
negligent) enf orc ement of such la ws, regardless of whether they are among those who believ e
these la ws to be just. But where it is not reasonab le to def end an unjust la w , citizens will not
ha v e a duty to share in the c ompensatory c osts of the enf orc ement of such la ws. State officials
do not enjo y immunity f or instanc es of u n reasona b le harm.
Some seriously unjust la ws are subje ct to reasonab le disagre ement; some mildly unjust la ws
are not. Be c ause the degre e of reasonability and the degre e of injustic e c ome ap art in this w a y ,
the abo v e c onsiderations imp ly that w e c an’t inf er whether an official is immune from the
enf orc ement c osts of some la w on the b asis of the degre e of the injustic e of the la w . Likewise,
the abo v e c onsiderations imp ly that state officials are not guarante e d immunity simp ly b y the
non-negligent e x e cution of their ro le. Citizens do not o w e it to state officials to share in the
c ompensatory c osts of their enf orc ement of unreasonab le la ws, ev en if those officials are not
in an y w a y to b lame f or their actions.
4.6 Conclusion
The Price of Duty Ac c ount tells us that state officials enjo y vic arious immunity with re-
spe ct to man y sorts of harm. M an y of the c osts impose d b y state officials should be share d
b y the citizens who are morally re quire d t o p articip ate in the po litic al proje ct of which those
24
Here I w on’t take a stand on the c ontentious question what this relev ant body of pub lic inf ormation is,
and on what c onstitutes a reasonab le position relativ e to that body of inf ormation.
139
c osts are a f allout. But not a l l of the c osts impose d b y state officials. There are e x c eptions.
W e note d f our. Official immunity does not e xtend to instanc es of incidental, negligent, rene-
gade, or unreasonab le harm. That the Price of Duty Account suggests these e x c eptions
is not just a f eature, but a virtue, of the ac c ount. These are p lausib le e x c eptions. The ac c ount
fills out the penumbra in our pre-the oretic judgments about the limits of official immunity in
a w a y that does not c onflict with our more determinate judgments about official immunity .
On this ac c ount, the mor a l immunity enjo y e d b y state officials loo ks quite diff erent from
the leg a l immunity they enjo y in most c ontemporary nations. In m y o wn c ountry of the
Unite d States, f or instanc e, legal immunity outruns moral immunity along a number of
fronts. U.S. la w grants “abso lute immunity” to a number of offic es. Judges, prose cutors,
legislatures, and c ertain members of the e x e cutiv e branch are almost nev er made to p a y c om-
pensation, ev en if they’v e acte d negligently or unla wfully .
25
M ost other officials enjo y “ qual-
ifie d immunity”, under which officials c an be made to p a y c ompensation f or the wrongful
harms they c ause, but only if they ha v e acte d in a w a y that an y “reasonab le person” w ould
kno w vio late d “ clearly-estab lishe d ” la w .
26
E v en this standard, ho w ev er, e xtends legal immu-
nity w ell bey ond the bounds of moral immunity , as c onduct c an easily impose incidental or
negligent damages without p assing the ‘ clearly-estab lishe d ’ and ‘reasonab le person’ tests.
27
And ev en where the la w does ho ld an official liab le to p a y c ompensation, man y such officials
25
Fre d Smith, “Loc al So v ereign Immunity ,” Co lu m bia Law Revi ew 116 (2016): 411.
26
Ha rlow v. Fitzger a ld , 457 U.S. 800, 818 (1982).
27
Courts ha v e generally interprete d the “ clearly-estab lishe d ” test as a very demanding one. “The Court has
said that, e x c ept in e xtraordinary circumstanc es, the la w is clearly estab lishe d only if a prior c ase has de clare d the
c onduct unc onstitutional. And that prior c ase must ha v e f acts that map neatly onto the f acts of the p laintiff’s
c ase” (Jo anna Sch w artz, “Qualifie d Immunity’s Bo ldest Lie,” Chic ago La w Review 88 no . 3 (2021): 607). Se e
also T yler Finn, “Qualifie d Immunity F ormalism: ‘Clearly Estab lishe d La w’ and the Right to Re c ord P o lic e
Activity ,” Co lumbia La w Review , 119 no . 2 (2019): 445-486; and Jo hn Jeffries, Jr., “The Liability Rule f or
Constitutional T orts,” V irginia La w Review , 99 no . 2 (2013): 207-270.
140
enjo y indemnific ation. This is espe cially true of la w enf orc ement offic ers, who , ev en when
they are not prote cte d b y qualifie d immunity , almost alw a ys ha v e their c ompensatory burden
assume d b y their dep artment or loc al/ state go v ernment.
28
This, of c ourse, is just an indire ct
w a y of shifting an offic er’s c ompensatory burden onto the taxp a y ers.
The f act that legal immunity in the Unite d States outruns moral immunity is not y et to
sa y that U.S. immunity po lic y is deficient. But it is to sa y that the po lic y is deficient u n less
there is a justific ation f or this div ergenc e. When the state grants legal immunity f or an offi-
cial ’s incidental, negligent, renegade, or unreasonab le harm, the state chooses not to enf orc e
a c ompensatory duty that is o w e d b y that official. This is pro tanto wrongful: the state has
a pro tanto duty to ensure the satisf action of the c ompensatory claims o f its citizens, and
this duty is all the more demanding in the c ase of official harm, as the state typic ally bears
some c omp licity f or the harm done b y its o wn agents. The upshot: the state had better ha v e
a c ompelling justific ation f or the w a ys in which it e xtends legal immunity bey ond the limits
of moral immunity .
It is a c omp le x and largely-empiric al question whether there is such a justific ation. Speak-
ing f or m yself, I’m skeptic al. The standard arguments in f a v or of such e xtensiv e immunity
regimes as w e ha v e in the Unite d S tates purport that a less e xtensiv e regime w ould deter offi-
cials from doing their jobs w ell, deter quality c andidates from app lying f or state offic es, and
tie up essential go v ernment w orkers and resourc es in c ourt.
29
But there is a gro wing c onsen-
sus amongst legal scho lars that these arguments are empiric ally untenab le, and that official
immunity c ould be sc ale d b ack in the Unite d States without serious o v erdeterrenc e or ad-
28
Jo anna “Sch w artz, P o lic e Indemnific ation,” NYU La w Review 89 no . 3 (2014).
29
Se e the SCO TUS opinion in Ha rlow v. Fitzger a ld , 457 U.S. 800, 814 (1982).
141
ministrativ e c osts.
30
P erhaps there are better arguments out there. But notic e the hurdle
these arguments ne e d to clear. It w ouldn’t be enough to sho w that the benefits of a more
e xp ansiv e immunity po lic y are greater than the benefits of a less e xtensiv e one (a challenge in
its o wn right). It must also be sho wn that this diff erenc e in benefit is so great as to outw eigh
the state’s pro tanto duty to enf orc e the c ompensatory duties of its officials.
30
This is largely due to the c omprehensiv e w ork of Jo anna Sch w artz. In addition to her “Indemnific ation”
and “Bo ldest Lie”, se e also “Ho w Qualifie d Immunity F ails,” Y a le Law Jou rna l 127 no . 2 (2017): 2-76; “The
Case Against Qualifie d Immunity ,” Notre Da me Law Revi ew 93 no . 5 (2018): 1797-1851; and “ After Qualifie d
Immunity ,” Co lu m bia Law Revi ew 120 no . 2 (2020): 309-383. F or a p artial def ense of qualifie d immunity
se e Aaron Nielson and Christop her W alker, “ A Qualifie d Def ense of Qualifie d Immunity ,” Notre Da me Law
Revi ew 93 no . 5 (2018): 1853-1885.
142
Ref erenc es
\parskip :12.0pt p lus 1.20007pt minus 2.39996pt and \parindent :0.0pt
Larry Ale x ander, “Re cipe f or a The ory of Self-Def ense,” in Christian Coons and Michael
W eber ( e ds.), The Et hi cs of Se lf- Defense (Oxf ord Univ ersity Press 2016).
Andrew Altman and Christop her Heath W ellman, “From Humanitarian Interv ention to
Assassination: Human Rights and P o litic al V io lenc e,” Et hi cs 118 (2008): 228-257.
Americ an La w Institute, M odel P enal Code (Philadelp hia: The Institute, 1985).
Americ an La w Institute, Restatement (Third) of Agenc y (St. P aul, Minnesota: Americ an
La w I nstitute Pub lishers, 2006).
Thomas Aquinas, Su m ma Theo logi ca (F athers of the Dominc an Pro vinc e trans., Benzinger
Bros. 1948).
Christian Barry and Robert Kirb y , “Sc epticism about Beneficiary P a ys: A Critique,” Jou rna l
of A pp li ed Phi losop hy 34 (2017): 282-300.
Sab a Bazargan-F orw ard, “Comp licitous Liability in W ar,” Phi losop hi ca l Stu di es 165 (2013):
177-195.
—, “Killing Minimally Responsib le Threats,” Et hi cs 125 (2014): 114-136.
—, “Def ensiv e Liability W ithout Culp ability ,” in Christian Coons and Michael W eber (e ds.),
The Et hi cs of Se lf- Defense (Oxf ord Univ ersity Press 2016).
143
—, “Grounding the Beneficiary P a ys Princip le,” Oxf ord Studies in P o litic al Philosop h y 8
(f orthc oming).
Josep h Bo w en, “Humanitarian Interv ention, Other-Def ense, and Consent,” Oxford Stu di es
i n P o liti ca l Phi losop hy , V o l. 8 (Oxf ord Univ ersity Press, f orthc oming).
Daniel Butt, “‘ A Doctrine Quite New and Altogether Untenab le’: Def ending the Beneficiary
P a ys P rincip le,” Jou rna l of A pp li ed Phi losop hy 31 (2014): 336-348.
Ale x andra Cuoto , “The Beneficiary P a ys Princip le and Strict Liability ,” Phi losop hi ca l Stu di es
175 (2018): 2169-2189.
Kaila Draper, “Ne c essity and Proportionality in Def ense,” in The Et hi cs of Se lf-Defense , e ds.
Christian Coons and Michael W eber (Oxf ord: Oxf ord Univ ersity Press, 2016).
—, “Def ense,” Phi losop hi ca l Stu di es 145 (2009): 69-88.
Cé cile F abre, “P ermissib le Rescue Killings,” Proc eedi ngs of t he A ristote lia n Soci ety 109 (2009):
149-164.
Joel F einberg, “V o luntary Euthanasia and the Inalienab le Right to Lif e,” Phi losop hy a nd Pu b-
li c Aff a i rs 7 (1978): 93-123.
Kimberly Kessler F erzan, “The Bluff: The P o w er of Insinc ere Actions,” Leg a l Theory 23
(2017): 168-202.
Christop her Finla y , “Legitimac y and Non-State P o litic al V io lenc e,” Jou rna l of Po liti ca l Phi-
losop hy 18 (20 10): 287-312.
T yler Finn, “Qualifie d Immunity F ormalism: ‘Clearly Estab lishe d La w’ and the Right to
Re c ord P o lic e Activity ,” Co lu m bia Law Revi ew , 119 (2019): 445-486.
Jo anna M ary Firth & Jonathan Quong, “Ne c essity , M oral Liability , and Def ensiv e Harm,”
Law a nd Phi losop hy 31 (2012): 673-701.
Ge orge Fletcher, Ret hi n ki ng Cri mi na l Law (Oxf ord: Oxf ord Univ ersity Press, 2000).
Helen Fro w e, Defensive Ki l li ng (Oxf ord: Oxf ord Univ ersity Press, 2014).
—, “Judging Arme d Humanitarian Interv ention,” in The Et hi cs of A rmed Hu ma nita ria n
Inteventi on , e d. Don E. Scheid (Cambridge, 2014).
144
Alberic o Gentili, De Ju re Be l li (On t he Law of W a r) , trans. Jo hn C. Ro lf e (Clarendon Press
1933)(1598).
Robert Goodin, “Disgorging the Fruits of Historic al W rongdoing,” A meri ca n Po liti ca l Sci-
enc e Revi ew 107 (2013): 478-491.
Kerah Gordon-So lmon, “What M akes a P erson Liab le to Def ensiv e Harm?,” Phi losop hy a nd
Phenomeno logi ca l Resea rc h 97 (2018): 543-567.
Kalle Grill, “Liberalism, Altruism, and Group Consent,” Pu b li c Hea lt h Et hi cs 2 (2009): 146–
157.
Ha rlow v. Fitzger a ld , 45 7 U.S. 800, 818 (1982).
H.L.A. Hart, “ Are There An y Natural Rights?” Phi losop hi ca l Revi ew 64 (1955): 175-191.
Jo hn Jeffries, Jr., “The Liability Rule f or Constitutional T orts,” V i rgi nia Law Revi ew , 99
(2013): 207-270.
René e Jorgensen, “The M oral Grounds of Reasonab ly Mistaken Self-Def ense,” Phi losop hy
a nd Phenomeno logi ca l Resea rc h 103 (2021): 140-156.
—, “M oral Risk and Communic ating Consent,” Phi losop hy a nd Pu b li c Aff a i rs 47 (2019):
179-207.
Seth Lazar, “Ne c essity in Self-Def ense and W ar,” Phi losop hy a nd Pu b li c Aff a i rs 40 (2012):
3-44.
—, “ Authorization and the M orality of W ar,” Austr a lasia n Jou rna l of Phi losop hy 94 (2016):
211-226.
Sigurd Lindstad, “Beneficiary P a ys and Respe ct f or Autonom y ,” Socia l Theory a nd Pr acti c e
47 (2012): 153-169.
Jeff M cM ahan, The Et hi cs of Ki l li ng: Pro b lems at t he Ma rgi ns of Life (Oxf ord Univ ersity
Press, 2002).
—, “The Basis of M oral Liability to Def ensiv e Killing,” Phi losop hi ca l Issu es 15 (2005): 386-
405.
—, Ki l li ng i n W a r (Oxf ord: Oxf ord Univ ersity Press, 2009).
145
—, “Humanitarian Interv ention, Consent, and Proportionality ,” in Et hi cs a nd Hu ma nity:
Themes from t he Phi losop hy of Jonat ha n Glover , e ds. N. Ann Da vis, Richard Keshen &
Jeff M cM ahan (Oxf ord, 2009).
—, “Who is M orally Liab le to be Kille d in W ar?” A na lysis 71 (2011): 544-559.
—, “What Rights M a y be Def ende d b y M eans of W ar?” The Mor a lity of Def ensive W a r , e ds.
Cé cile F abre and Seth Lazar (Oxf ord: Oxf ord Univ ersity Press, 2014).
—, “The Limits of Self-Def ense,” in The Et hi cs of Se lf-Defense , e ds. Christian Coons and
Michael W eber (Oxf ord: Oxf ord Univ ersity Press, 2016).
—, “Liability , Proportionality , and the Number of Aggressors ”, in The Et hi cs of W a r , e ds.
Sab a Bazargan and Samuel Rickless (New Y ork: Oxf ord Univ ersity Press, 2017).
—, “Ne c essity and Proportionality in M orality and La w ,” in Nec essity a nd Proporti ona lity i n
Internati ona l Peac e a nd Secu rity Law , e ds. Claus Kres and Robert La wless (New Y ork:
Oxf ord Univ ersity Press, 2021).
Aaron Nielson and Christop her W alker, “ A Qualifie d Def ense of Qualifie d Immunity ,” Notre
Da me Law Revi ew 93 (2018): 1853-1885.
Robert Nozick, A na rc hy, State, a nd Utopia (W iley-Blackw ell, 2001).
Michael Otsuka, “Killing the Innoc ent in Self-Def ense,” Phi losop hy a nd Pu b li c Aff a i rs 23
(1994):74-94.
—, “The M oral Responsibility Ac c ount of Liability to Def ensiv e Killing”, in The Et hi cs
of Se lf-Defense , e ds. Christian Coons and Michael W eber (Oxf ord: Oxf ord Univ ersity
Press, 2016).
Derek P arfit, On What Matters , v o l. 1 (Oxf ord, 2011).
T om P arr, “The M oral T ainte dness of Benefiting from Injustic e,” Et hi ca l Theory a nd Mor a l
Pr acti c e 19 (2016): 985-996.
Jonathan P arry , “Def ensiv e Harm, Consent, and Interv ention,” Phi losop hy a nd Pu b li c Af-
f a i rs 45 (2017): 356-396.
Jonathan Quong, The Mor a lity of Defensive Forc e (Oxf ord: Oxf ord Univ ersity Press, 2020).
Jo hn Ra wls, A Theory of Justi c e (Cambridge, M assachusetts: Harv ard Univ ersity Press, 1971).
146
Josep h Raz, “Responsibility and the Negligenc e Standard,” Oxford Jou rna l of Leg a l Stu di es
30 ( 2010): 1-18.
Da vid Rodin, W a r a nd Se lf-Defense (Oxf ord: Oxf ord Univ ersity Press, 2002).
—, “Justifying Harm,” Et hi cs 122 (2011): 74-110.
Jo anna Sch w artz, “Qualifie d Immunity’s Bo ldest Lie,” Chi cago Law Revi ew 88 (2021): 605-
684.
—, “P o lic e Indemnific ation,” NYU Law Revi ew 89 (2014): 885-1005.
—, “Ho w Qualifie d Immunity F ails,” Y a le Law Jou rna l 127 (2017): 2-76.
—, “The Case Against Qualifie d Immunity ,” Notre Da me Law Revi ew 93 (2018): 1797-
1851.
—, “ After Qualifie d Immunity ,” Co lu m bia Law Revi ew 120 (2020): 309-383.
A. Jo hn Simmons, “The Princip le of F air Pla y ,” Phi losop hy a nd Pu b li c Aff a i rs 8 (1979): 307-
337.
Sc ott Shapiro , Leg a lity (Harv ard, M assachusetts: Harv ard Univ ersity Press, 2011).
Fre d Smith, “Loc al So v ereign Immunity ,” Co lu m bia Law Revi ew 116 (2016): 409-488.
V ictor T adros, The Ends of Ha rm: The Mor a l Fou ndati ons of Cri mi na l Law (Oxf ord: Ox-
f ord U niv ersity Press, 2011).
—, “Duty and Liability ,” Uti litas 24 (2012): 259-277.
—, W rongs a nd Cri mes (Oxf ord: Oxf ord Univ ersity Press, 2016).
—, “Causation, Culp ability , and Liability ,” in The Et hi cs of Se lf-Defense , e ds. Christian
Coons and Michael W eber (Oxf ord: Oxf ord Univ ersity Press, 2016).
—, “Orw ell ’s Battle with Brittain: V ic arious Liability f or Unjust Aggression,” Phi losop hy a nd
Pu b li c Aff a i rs 42 (2014): 42-77.
F ernando T esón, Hu ma nita ria n Interventi on , 3rd e d. (Ardsley , NY : T ransnational, 2005).
Richard Thaler and Cass Sustein, “Libertarian P aternalism,” The A meri ca n Ec onomi c Revi ew
93 (2003): 175–179.
147
Judith Jarvis Thomson, “Self-Def ense,” Phi losop hy a nd Pu b li c Aff a i rs 20 (1991): 283-310.
—, “Pref erential Hiring,” Phi losop hy a nd Pu b li c Aff a i rs 2 (1973): 364-384.
Suzanne Uniacke, “Proportionality and Self-Def ense,” Law a nd Phi losop hy 30 (2011): 253-
272.
Daniel V iehoff, “Legitimac y as a Right to Err,” in NOMOS LXI: Po liti ca l Legiti macy , e ds.
Jack Knight and M elissa Sch w artzberg (New Y ork: NYU Press, 2019).
Gary W atson, “Raz on Responsibility ,” Cri mi na l Law a nd Phi losop hy 10 ( 2016): 295-409.
148

Abstract (if available)

Abstract Responding to Harm is a collection of essays concerning the ethics of harm. The first three chapters concern the subject of defensive or preventative harm — that is, harm that is imposed on someone to prevent harm to someone else. The final chapter concerns compensatory morality; it concerns the nature of our duties to pay compensation for harms that have already occurred. All four chapters explicate, in different ways, the nature and moral significance of our duties to assume costs from others.

Chapter 1: “The Demands of Necessity”. It is widely held that defensive harm must be both necessary and proportionate. You shouldn’t kill someone in self-defense who is only trying to pinch you on the arm; that would be disproportionate. But even when an attacker poses a lethal threat, you shouldn’t kill them in self-defense if you can just as well stop their attack with a slap on the wrist; killing them would then be unnecessary. In chapter one I develop a novel explanation, precisification, and unification of the proportionality and necessity constraints. What seem to be two independent constraints, I argue, are not. Unjust attackers have a duty to avert their own attack. This duty could require them to bear up to a certain amount of cost to avert their attack were they able. And this in turn determines what others may do to them: an attacker may only be made to bear as much cost to avert their attack as they could be required to take upon themselves. The proportionality and necessity constraints, I argue, both express this fundamental principle, only at different levels of generality.

Chapter 2: “Refusing Protection”. The second chapter takes up a different question about the limits of defensive harm. Just as we can often use force to defend ourselves, so too can we often use force to defend others. But suppose someone doesn’t want our protection. A committed pacifist asks you not to use violence to protect her. A powerful aggressor nation invades a weaker neighbor, but the weaker nation votes against welcoming the military intervention of the United States. What then? I argue that such refusal sometimes makes it wrong to intervene. This power is an upshot of a more fundamental power we each have to alter the “enforceability” of our rights. It’s familiar that we have powers to transfer or waive rights. For the same reasons that we have these powers over our rights, I argue, we have the power to change whether a right may be enforced, and by whom. One way in which this power can be exercised is by the refusal of protection: by refusing my protection you remove me from the class of persons with respect to whom your attacker is liable, thereby making it so that your attacker is not liable to defensive harm at my hands. This chapter defends the idea that we have such powers over our rights, and explores the implications of this idea for unwanted military and police interventions.

Chapter 3: “Defense Without Threat”. This chapter combines and develops some of the ideas from the first two chapters to answer an important challenge for the ethics of defense. Consider a paradigmatic case of self-defense. A villain attempts to murder you, and you’ll be killed unless you kill him first. The villain is clearly liable to be killed by attacking you — that is, he lacks his normal right that you not kill him, which largely explains why you are permitted to kill him.
The villain’s liability sure seems to have something to do with the fact that he would harm someone if he isn’t harmed him first. That seems, at first glance, at least a necessary condition for liability. But there’s a problem. Suppose a hitman attempts to kill you, but unbeknownst to you their gun is jammed. Or suppose a bank robber ”bluffs” that he will kill you if you don’t open the vault. You don’t need to harm the hitman or the bank robber to prevent yourself from being harmed. But it sure seems they are nonetheless liable to be harmed, unaware as you that they pose no actual threat. Our theory of liability needs to speak to such cases. I call this the Challenge of Merely Apparent Attackers.
Chapter three presents an unorthodox answer to this challenge. I develop and defend a theory of liability according to which merely apparent attackers are not strictly liable to be harmed. Their rights against harm persist, and those rights continue to give us reason not to harm them. Nonetheless, characters like the hitman and the bank robber have a duty to assume the costs of the defensive actions that are taken against them, and this duty explains why they lack many of the usual perks or upshots of their rights. They are what I call quasi-liable. Their moral status is neither that of an innocent person nor that of an attacker who poses a genuine threat of harm.

Chapter 4: “The Price of Duty”. State officials regularly impose harm on the citizens they are supposed to serve, some of it wrongful. Who should bear the burden of paying compensation for these wrongs? Should it be the agents themselves, or should the burden be spread across the citizenry in the form of taxation? In chapter four I defend a theory of limited official immunity, according to which the citizenry should assume the costs of certain official wrongs. I argue that limited official immunity is an upshot of a general principle of distributive justice, according to which those who are morally required to participate in some project are required to share certain costs of that project. Understanding official immunity as derivative of this principle helps us to identify, not just the grounds, but also the limits of official immunity: it helps us to sort between the kinds of wrongs for which the burdens of official wrongs should fall on the individual agents themselves and the kinds of wrongs for which those burdens should be spread across the citizenry.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The causal account of liability to self-defensive harm reconsidered

PDF

The virtue of reasonableness: on the normative epistemic implications of public reason liberalism

PDF

Getting our act(s) together: the normativity of diachronic choice

PDF

On law enforcement officers’ liability to defensive harm

PDF

On intimacy: a philosophical account

PDF

Responsibility and rights: a search for a principled distinction between criminal law and tort law

PDF

Mistaken defense and normative conventions

PDF

Aggregating complaints

PDF

Public justification beyond legitimacy

PDF

Rationality and the primacy of the occurrent

PDF

Aggregating happiness: seeking and identifying a single plausible unifying theory

PDF

A deontological explanation of accessibilism

PDF

Contrastive reasons

PDF

What "ought" ought to mean

PDF

The Internet middlemen: targeting intermediary firms as gatekeepers in the online economy

PDF

Defending industrial control systems: an end-to-end approach for managing cyber-physical risk

PDF

Identifying and mitigating safety risks in language models

PDF

The constitution of action

PDF

Improving binary program analysis to enhance the security of modern software systems

PDF

Leveraging programmability and machine learning for distributed network management to improve security and performance

Asset Metadata

Creator Clark, David James (author)

Core Title Responding to harm

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Philosophy

Degree Conferral Date 2023-08

Publication Date 07/21/2023

Defense Date 05/09/2023

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

Tag attackers,Compensation,consent,culpability,Defense,distributive justice,duties,ethics,harm,immunity,liability,Morality,necessity,OAI-PMH Harvest,philosophy,policing,proportionality,Protection,reasons,refusal,responsibility,rights,War

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Quong, Jonathan (committee chair), Schroeder, Mark (committee chair), Keating, Greg (committee member), Nebel, Jacob (committee member), Ross, Jacob (committee member)

Creator Email clarkdj@usc.edu,davidjclark21@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113282112

Unique identifier UC113282112

Identifier etd-ClarkDavid-12125.pdf (filename)

Legacy Identifier etd-ClarkDavid-12125

Document Type Dissertation

Format theses (aat)

Rights Clark, David James

Internet Media Type application/pdf

Type texts

Source 20230721-usctheses-batch-1072 (batch), 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.

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