. 3
( 10)


be killed. Still, you should not wave your wand, even though it would
maximize earthy goods like pleasure, happiness, and desire satisfaction to
do so, even though it would maximize the number of lives and minimize
the number of murders.

MAC provides the resources for an appropriate response to both aspects
of this objection. We will begin with the ¬rst aspect “ the moral signi¬-
cance of life, health, and the like “ though our discussion will inevitably
creep toward the second. And we will begin abstractly, but move toward
a relatively concrete treatment of two cases: our duty not to lie and our
duty not to murder.
The three main points I will make are simply stated. First, the relevant
intuitive goods are indirectly supported by MAC as necessary precondi-
tions for normatively proper actions. One™s role in certain weighty group
acts, and indeed the very existence of such acts, will sometimes require
respect for these goods “ for instance, informed belief “ even if the basic
normative principle conjoined with MAC is hedonistic in the manner
of HMP or classical utilitarianism. MAC will deliver intuitive goods of
these sorts even if the only basic normative good is pleasure. Second,
MAC can deliver an appropriate account of intuitive normative distinc-
tions between harming and not aiding, which underlie some standard
deontological restrictions on action. Third, there are certain normative
practices that are weighty group acts in my sense and that forbid certain
sorts of activity by participating atomic agents, and these support intuitive
deontological constraints. For instance, our normative practice of gener-
ally not lying is a group act that is quite valuable, and so VLD and DD
will seldom countenance defecting lies.
This and the following two sections will elaborate these three points
sequentially, and work toward a relatively detailed treatment of our duties
not to lie and murder. In Chapter 8, this general conception will be
elaborated in a code that delivers all standard deontic duties.
First point: Standard forms of consequentialism do not assure in the
proper way the basic normative signi¬cance of certain sorts of goods.
Life, true belief or knowledge, ordinary physical and mental abilities,
freedom from injury, and freedom from manipulation and imprison-
ment seem normatively signi¬cant in a very deep way. Indeed, they
seem very signi¬cant even if someone™s happiness is served by their
own death, false belief, injury, disability, imprisonment, or manipu-
lation. But Multiple-Act Consequentialism helps in this regard, even
when conjoined with a classical hedonistic or desire-satisfaction value
Notice that true beliefs, life, physical and mental abilities, freedom from
injury, and freedom from imprisonment and manipulation are standardly
prerequisite for successful agency. Hence, as a mnemonic and abbrevia-
tion, we may call these goods “agency goods”. Notice also that among

the reasons whose acceptance MAC suggests are some that support or
constitute various sorts of agency goods.
The ¬rst class of these are among the agent-constituting reasons that
constitute conditions for group agency, and hence foster life, freedom from
injury and imprisonment, and true belief, both for one™s future selves
and for other people within group agents in which one partakes. The
dead, severely injured, imprisoned, and misled cannot effectively coop-
erate toward group goals. Agent-constituting reasons that foster agency
goods within group agents can be of two sorts. Some support intuitively
positive action in support of the goods. But some are negative restrictions
against murder, lying, and injuring within the group.
And notice that there are general group actions on the part of at least
most humans “ or, at the very least, most human noncombatants “ that
are forms of one-off cooperation speci¬cally eschewing murder, injury,
and the like, at least in many circumstances, and that have positive conse-
quences in even a classically utilitarian or otherwise hedonist sense. Most
of us are parts of one-off group agents in my sense with such appropriate
normative projects. That is a very big fact for MAC, to which we will
return shortly.
This general cooperative activity reinforces a second class of relevant
reasons that MAC supports. Other agents can effectively act toward goals
only when their agency goods are respected by us, from the outside. So
we can expect that MAC will stipulate that other agents with appropriate
projects also involve agency goods that require our support. But even
other agents without appropriate projects will in some degree reap the
bene¬ts of our own general group acts of refraining from murder and the
Indeed, MAC supports agency goods in so many ways that it may seem
to threaten double- or triple-counting of such goods. But the weight of
agency goods in MAC is felt only indirectly through the mechanism
of VLD and DD. Agency goods do not feed directly into assessment of
consequences, excepting some complications to which Chapter 8 will
return, nor are we to engage in any summation of their value as delivered
by various group practices.
MAC also suggests other characteristic complications. For one thing,
the projects of agents must be appropriate if their agency is properly to
command moral respect in some of the ways I have noted. They must
be proper projects in the sense introduced in the last chapter. In gen-
eral, agency goods are only genuine agency goods when they support
appropriately bene¬cent projects. This may seem to provide intuitive but

“nongenuine” agency goods with insuf¬cient protection. For instance,
while the dead can™t act, they of course can™t do wrong either. Still, mutual
restraint from mutual murder is one very weighty form of group action
according to MAC, and it seems a quite minimal level of cooperation
required for any effective human group action. And by the strictures of
that group act, we restrain ourselves at least to some degree from mur-
dering even those humans who are outside the group act or who defect
from its requirements.
Other agency goods, beyond life, require somewhat different and more
complicated treatments, both intuitively and according to MAC. True
belief is perhaps a paradigmatic case. Let me merely list some of the
complications: (i) Those with false beliefs cannot characteristically act
effectively. (ii) And there is a one-off cooperative practice of not telling
at least explicit lies in many circumstances that has signi¬cant normative
weight, and is a group agent. (iii) But the way in which circumstances
are relevant according to that cooperative practice is complicated and
indeed somewhat indeterminate. Our practice clearly allows us to lie
sometimes, and sometimes does not determinately either forbid or allow
lying. (iv) There is also a ¬nite limit to the normative weight of that prac-
tice according to MAC. (v) Independent of that particular general group
act forbidding lying, MAC suggests that true belief will have special nor-
mative signi¬cance only when it somehow supports the effective pursuit
of proper goals. (vi) MAC also recommends lying within a group agent
with a consequentially harmful project, if that harmful project overbal-
ances the normative weight of our general group act eschewing lying or
if that general act doesn™t forbid that particular sort of lie. (vii) False belief
only characteristically undercuts effective action. Sometimes group activity
toward a proper project will be furthered by lying of some sorts within
the group. (viii) And our one-off group act eschewing explicit lying is
not the only general form of cooperative practice regarding truth telling
that is possible. Perhaps a general group act that required positive truth
telling in support of proper projects, and also forbade misleading in similar
circumstances, would be better, and it may also be a form of group action
of at least limited extent that one can foster by more or less prospective
cooperation. (ix) And prospective cooperation may be very weighty on
the timeless conception of relevant group agents implied by the treatment
of continuing agency as group agency. (x) Another complexity is that,
even independent of the proper forms of our one-off general group acts
regarding lying and truth telling, the positive demands of true belief go
beyond an absence of lies, especially within certain forms of small-scale

group agent with weighty proper projects and in support of the proper
projects of other agents.
This is a complex set of analytical factors. We will return in section III
to a more concrete discussion of our obligations regarding lying according
to MAC. But the immediate point is that MAC requires a relatively com-
plicated story about when apparent agency goods like life and true belief
require moral respect. They are genuine agency goods when in support of
appropriate projects, at least when the proper projects in question are not
outweighed by other projects. And they also properly command moral
respect when they are otherwise protected by general group acts with
proper projects. And agency goods can be fostered by accepted reasons
recommended by MAC in a variety of different ways “ for instance, by
negative restrictions on action as well as by positive exhortations. And
there are also the various complications I have just noted for the case of
true belief.
This general picture is in accord with any determinate strictures of
our re¬‚ective common sense, I believe. MAC grants truth and life and
other agency goods a more or less fundamental normative status, but a
more complex and nuanced status than some traditional theories allow.
And these complexities are intuitive, as we will shortly see in suitable
Another relevant intuitive complexity of MAC is this: MAC suggests a
two-tiered account of what I will call fundamental normative goods. At the
¬rst and most basic level is whatever value is recognized in the basic con-
sequentialist principle with which it is allied. It might be pleasure or desire
satisfaction. But there is no action in accord with a basic principle unless
there are agents. And effective agency in accord with a basic normative
principle requires agency goods. Genuine agency goods are fundamental
according to MAC, even if not quite basic. So too are other agency goods
protected by bene¬cent one-off moral practices that are group agents.
Some think that we ought ¬rst to maximize agency goods and only then
turn to a secondary maximization of familiar consequentialist goods like
pleasure and pain. Something like this has been suggested, for instance,
by David Cummiskey.1 But MAC instead grants a kind of deontological
status to respect for certain agency goods within group agents and even
among distinct agents, at least when the agency goods are in the proper

1 Cummiskey (1996).

service of basic value, and also grants other forms of nearly deontic respect
to fundamental agency goods protected by weighty group acts that we are
inside. It organizes two intuitive types of goods in a somewhat different
but still plausibly intuitive way.
Another way to look at this structure of fundamental goods is by com-
parison to both standard utilitarian and Kantian treatments of nonhuman
animals. Nonhuman animals can feel pain, but for the most part coop-
eration with nonhuman animals is impossible. So standard utilitarianism
suggests that the standing of nonhuman animals as moral patients is basi-
cally the same as that of humans. But standard Kantian or contractarian
views suggest that they have no normative standing. MAC allows some-
thing more complex, and indeed more intuitive, at least when allied with
a classical value theory. Nonhuman animals have moral standing of indeed
the primary and basic sort. Depending on the details of the bene¬cent
group acts of humans, they may even have a further derived status as moral
patients that should engage nearly deontic respect from those humans. But
humans who engage in proper group action, or indeed individual activity,
have a second sort of fundamental but not basic normative status, which
indeed makes them moral patients in another but more exalted sense. They
can lose some of this status by the viciousness of their projects. They can
lose the respect due to proper cooperators and bene¬cent agents. That
respect is not unconditional respect. Still, no matter how vicious they
are, they retain the basic normative status of any sentient animal, and
that is a real normative status, which enjoins a real though different form
of unconditional respect. The picture of fundamental normative goods
suggested by MAC is somewhat complicated, but it is complicated in an
intuitive way.
In sketching the forms of respect that agency goods deserve according
to MAC, we have edged toward an answer to the second aspect of our
objection. MAC suggests a nearly deontological respect for agency goods
within forms of group agents with appropriate projects, and also among
distinct agents with proper projects. It may even suggest a nearly deonto-
logical respect by us for these goods wherever they are found, depending
on the details of our own general moral practices that are group acts.
Indeed, there is a degree to which it is arbitrary whether we complicate
the account of agency goods, or instead complicate the account of roughly
deontological respect that putative agency goods deserve. But let me now
focus our discussion more closely and directly on that second aspect of
our standing objection.


The second and perhaps most characteristic aspect of our ¬rst standard
objection to act consequentialism re¬‚ects the inclination of common-
sense morality toward what Samuel Schef¬‚er has called “agent-centered
restrictions”.2 Agent-centered restrictions forbid particular agents from
doing certain things, even if the cost of their failure to do such a thing is
that many other acts of exactly the same sort will be performed instead by
others. For instance, an act that tortures the innocent may seem intuitively
forbidden. It may seem intuitively forbidden even if the certain cost of
refraining from that act of torture will be that many other people commit
many such tortures.
Some may object that such situations are not coherently described,
because to fail to perform the ¬rst act of torture is in fact to commit
the others. But common sense incorporates a series of normative distinc-
tions between acting and refraining, doing and allowing, harming and
letting be harmed, which may be deployed to defend this aspect of the
objection against that resistance, and which play other crucial roles in
our commonsense morality. They are our focus in this section. We will
complete our discussion of agent-centered restrictions only in the next
My argumentative situation is delicate at this point. In the end, as
will be seen in Chapter 7, I will not admit that there is a deep distinction
between acting and refraining, nor between omitting and committing.
Consequentialism, in all its direct forms, inevitably avoids such a deep
distinction. And yet I must deliver some normative differences between
these things if consequentialism is to be suitably intuitive. MAC can deliver
what is required, not too much and not too little, though it will take a
little time to see this properly.
One way we can study the interaction of MAC and these distinctions
is to grab a ride on the Trolley Problem. And so into the belly of the
deontological beast we go.
The fashion for philosophy rooted in intuitions about concrete cases
has characteristic dangers. To focus on a few cases can mislead, since other
cases may push in different directions. And the absence of a systematic
account can hide crucial dif¬culties. Nevertheless, it is such cases that will
be our immediate concern.

2 Schef¬‚er (1982).

To get a sense of how MAC can help with the Trolley Problem, we will
begin with a quick review of what it ideally should deliver. The literature
on trolleys began with Philippa Foot™s classic “The Problem of Abortion
and the Doctrine of Double Effect.”3 But I will emphasize the also classic
development of Foot™s cases by Judith Jarvis Thomson in “Killing, Letting
Die, and the Trolley Problem” and “The Trolley Problem.”4 There are a
number of concrete examples to which we will need to attend in order to
collect a set of relevant intuitions. The puzzles of the trolley literature arise
from clanging together the diverse intuitions generated by these various
First, there is the case of Alfred and Bert. It is the standard assumption
within the trolley literature that there isn™t always a morally relevant dif-
ference between otherwise identical cases of killing and letting die, and
that this case shows that.
[I]magine that
(1) Alfred hates his wife and wants her dead. He puts cleaning ¬‚uid in her coffee,
thereby killing her, and that
(2) Bert hates his wife and wants her dead. She puts cleaning ¬‚uid in her coffee
(being muddled, thinking it™s cream). Bert happens to have the antidote to
cleaning ¬‚uid, but he does not give it to her; he lets her die.5

Since these actions seem equally abhorrent, it seems that we should con-
clude with Thomson that the difference between killing and letting die
isn™t always morally signi¬cant, which is of course not at all to conclude
that it isn™t sometimes quite signi¬cant.
Second, there is Foot™s original Trolley case, and its slight but signi¬cant
modi¬cation in Thomson™s Bystander case. Here is Thomson™s statement
of the original Trolley:
Suppose you are the driver of a trolley. The trolley rounds a bend, and there come
into view ahead ¬ve track workmen, who have been repairing the track. . . . [Y]ou
must stop the trolley if you are to avoid running the men down. . . . [T]he breaks
don™t work. . . . Now you suddenly see a spur of track leading off to the right. You
can turn the trolley onto it, and thus save the ¬ve men. . . . Unfortunately, . . . there
is one track workman on that spur of track. He can no more get off the track in

3 Foot (1967). This is reprinted in Foot (1978).
4 Thomson (1976, 1985). Both articles, as well as Foot (1967), are reprinted in Fischer and
Ravizza (1992).
5 Thomson (1976), in Fischer and Ravizza (1992: 69).

time than the ¬ve can, so you will kill him if you turn the trolley onto him. Is it
morally permissible for you to turn the trolley?6

With the majority, I intuit that this is morally permissible. Some think
that this is a matter of choice between killing one and killing ¬ve, since
as the driver of the trolley you are in charge of a hurtling object that will
certainly kill someone. Some disagree. So, for clarity, let™s introduce the
slight modi¬cations that give us the Bystander case:
[Y]ou have been strolling by the trolley track, and you can see the situation at
a glance: The driver saw the ¬ve on the track ahead, he stamped on the brakes,
the brakes failed, so he fainted. What do you do? Well, here is the switch, which
you can throw, thereby turning the trolley yourself. Of course you will kill one
if you do.7

With Thomson and the majority, I intuit that it is morally permissible
for you to turn the trolley. Hence sometimes it is morally permissible to
kill one to save ¬ve. Indeed, though this is apparently not the consensus
view, it seems to me to be morally mandatory in this case, though I would
hesitate to blame someone who failed to turn the trolley, because of the
moral dif¬culty of the situation.
But compare our third case, Transplant:
[Y]ou transplant organs, and you are such a great surgeon that the organs you trans-
plant always take. At the moment you have ¬ve patients who need organs. . . . If
they do not get those organs today, they will all die; if you ¬nd organs for them
today, they will all live. . . . [A] young man who has just come into your clinic for
his yearly check-up has exactly the right blood-type. . . . Lo, you have a possible
donor. All you need do is cut him up and distribute his parts among the ¬ve who
need them. You ask, but he says . . . no. Would it be morally permissible for you
to operate anyway?8

With the majority, I intuit that this would be morally impermissible. It is
not always morally permissible to kill one even to save ¬ve.
To underline this fact, consider a fourth case, Fat Man:
[Y]ou are standing on a footbridge over the trolley track. You can see a trolley
hurtling down the track, out of control. . . . [T]here are ¬ve workmen on the
track. . . . [Y]ou know of one certain way to stop an out-of-control trolley: Drop
a really heavy weight in its path. . . . It just so happens that standing next to you

6 Thomson (1985), in Fischer and Ravizza (1992: 280).
7 Ibid., 281.
8 Ibid., 280“281.

on the footbridge is a fat man, a really fat man. . . . [A]ll you need to do is give
him a little shove.9

Here again I have the intuition, shared by almost all, that this would be
So we have one case, Albert and Bert, which shows that sometimes
there is no morally signi¬cant difference between killing and letting die,
and a second case, Bystander, in which it seems at least permissible to kill
one to save ¬ve. But we have two other cases, Transplant and Fat Man,
in which it seems impermissible to kill one even to save ¬ve.
Consider two more cases, which introduce other factors to which we
will return. In reaction to the ¬rst new case, which I will call Bombing,
many intuit that it is morally permissible to bomb an enemy factory while
foreseeing but not intending that noncombatants will die, as a matter of
collateral damage. But then consider Hospital Gas: “Suppose . . . that there
are ¬ve patients in a hospital whose lives could be saved by the manufacture
of a certain gas, but that this inevitably releases lethal fumes into the room
of another patient whom for some reason we are unable to move.”10 In
this case, with the majority, I intuit that this action would be wrong,
despite its abstract similarity to Bombing.
There is a large literature that attempts to develop general explanations
that support our intuitive normative reactions to this range of cases. But
none of the obvious explanations generates intuitive results across the
board. If negative rights are sometimes intuitively stronger than positive
rights, still that difference doesn™t seem relevant in the Bert and Albert
case, and it apparently doesn™t differentiate Bombing from Hospital Gas. If
sometimes omissions that fail to save lives seem preferable to commissions
that take lives, still one properly acts in the Bystander case to kill one,
and that also doesn™t explain the moral symmetry of Bert and Albert.
If sometimes we have limited responsibility for foreseen but unintended
consequences of our actions, that doesn™t explain our reaction to Hospital
Gas. If sometimes the temporal order of a harm and a causally connected
good matters, that doesn™t explain our differential reactions to Hospital
Gas and Bombing.
Multiple-Act Consequentialism provides a new kind of general expla-
nation of our various responses to these various cases. It is what might be
called a framework explanation, which suggests the normative relevance

9 Ibid., 288.
10 Foot (1967), in Fischer and Ravizza (1992: 65).

of a variety of different factors in different cases. I believe it is in a sense
a deeper level of explanation than those we have just canvassed, which
explains why considerations such as those are important in some cases but
not in others.
Still, it would be wrong to claim too much for MAC in this context.
It is to a weak degree revisionary. It requires that not all our normative
intuitions about cases be probative, though we will see that it does allow
for the intuitive normative differences introduced through our survey of
famous concrete cases. This slightly revisionary nature seems unobjection-
able, since we cannot provide a coherent account that requires no revision
at all in our intuitions about all trolley cases.
Why is this? For one thing, the order in which people are presented
with concrete cases can affect the moral intuitions that the cases generate.
People seem to remember their initial response, and to try to preserve
consistency in their responses.11 Second, Unger has shown that cases
that involve four options can yield an intuitive response that is plausibly
inconsistent with the intuitive response to analogous two-option cases
that involve merely the deletion of two options from the ¬rst.12 Third,
there are also other groups of cases that show together that sometimes
something is wrong with our intuitions. Let me explain this third point.
Imagine a case like Fat Man, but in which the trolley is merely slowed
after running him over, and hence slowly comes to a stop before the ¬ve
workmen stuck on the tracks. Still, we retain our intuition that pushing
him onto the tracks to save the ¬ve is wrong. Now imagine instead that
the fat man is tied to a side track, which runs back onto the main line
before the point at which the ¬ve are stuck on the main track. Our choice
as a bystander is between letting the trolley run on straight to kill the ¬ve,
or turning it onto the side track, which will slow it down suf¬ciently
when it runs over the fat man to save the ¬ve, even though the trolley
will come back onto the main track and stop just before them. At least
if that case is presented after the original modi¬cation of Fat Man, most
intuit that this is also wrong. Because we will run into this case again,
I will give it a name, Sidetrack. Now imagine a third case, Loop. This
is a lot like Sidetrack. The track that is off to the side, with the fat man
tied down, continues on beyond where he is tied to the tracks. But rather

11 Unger (1996: 92).
12 Ibid., 88“94. Perhaps this is due to the fact that cognitive processing of complex cases involves
a certain sort of abstraction and simpli¬cation, at least when that effect isn™t swamped by
the order of presentation of cases.

than coming back onto the main line immediately, as in Sidetrack, it
instead loops around and comes back to the main line behind our ¬ve
workmen.13 If we do nothing, the ¬ve will be run over directly. If we turn
the trolley, it will be slowed by running over the fat man before it gets
back to our ¬ve, though it will come at them from behind, and would
have run them over from that direction if it hadn™t run him over. At least
if presented this case after the preceding cases, most people intuit that it is
wrong to turn the trolley onto the sidetrack, because the difference in the
way in which the track is arranged in Sidetrack and Loop seems trivial and
irrelevant. But notice that Loop is quite like Bystander, at least if there is a
fat man on the sidetrack in that case, but with the slight addition in Loop
of some unused track connecting the track running beyond the fat man
back to the main line. It is hard to believe that unused track could be of
grave moral signi¬cance, but in Bystander most intuit that it is permissible
to turn the trolley. These cases together suggest that we cannot coherently
treat all our intuitions about trolley cases as probative and stable.
Despite the fact that some revision of our intuitions about cases is
inevitable and desirable, still Multiple-Act Consequentialism provides a
mechanism to explain the intuitive differences among the various standard
cases with which we began this section. In essence, MAC suggests a range
of small factors that are sometimes relevant, and may sum together between
various cases in various ways to deliver intuitive results. To develop this
conception, ¬rst let me note some kinds of factors that MAC suggests
may be relevant. Then I will go on to apply this analytical framework to
deliver some of the more dif¬cult intuitive differences among the standard
cases with which we began.
Possible factors: First, when there is a particular group agent present in
a case, that may cause a difference in the proper normative assessment of
such a case according to MAC. For instance, it might allow us to de¬‚ect
a risk within a cooperating group agent but not onto an outsider. Or,
alternatively, it might be specially important to de¬‚ect the risk outside
the group in order to defend group cooperation. Notice that people tied
together on tracks might come for that very reason to share accepted
reasons that constitute them as a group agent of some sort, for though
their group attempts might be ineffectual, still they might be disposed to
cooperative activity. Or, alternatively, it might be that because they are
tied down, the individuals are in some ways not cooperating agents and

13 Loops like this were introduced by Thomson (1985). See Fischer and Ravizza (1992: 285).

hence not subject to the customary level of normative protection that that
Second, it might be that the exact form of proper treatment of the
de¬‚ection of risks and the like depends on a general cooperative practice
about such things that is itself a group agent, a group agent that constitutes
a general manner of handling the de¬‚ection of risks. If we all agree that
people shouldn™t be thrown onto the tracks when in the position of the
fat man in the original Fat Man case, if we have a one-off cooperative
practice that forbids that, that will be normatively relevant according to
MAC when the practice meets the conditions, noted in the last chapter,
required for group agency, and when it is a proper project.14 Deontologists
may not be happy that the propriety of such constraints is dependent on
our actual practices. But deontological intuitions formed within contrary
practices that were also suitably bene¬cent would plausibly be contrary
to intuitions formed within our own.
Third, temporal order is tied up with the existence of group agents in
interesting ways. Perhaps only those who precede an event can cooperate
to prevent it.
Fourth, there are obvious differences between the forms of group
agency that link people who share a hospital as doctors and patients and
those that link or fail to link different populations at war.
While, as I™ve said, MAC provides a mere framework explanation of
trolley cases, in which speci¬c group acts suggest the importance of dif-
ferent sorts of factors in different cases, it certainly suggests the gen-
eral overall importance of groups. That in itself may seem like another
general explanation of our intuitions, on a par with the distinction
between commission and omission or the Doctrine of Double Effect.
And indeed, both Robert Hanna and Peter Unger have suggested that
at least many of our different intuitions about these concrete cases can
be generally explained by reference to our treating various groupings of
people as normatively salient. We intuitively distinguish between inno-
cent bystanders “ like the fat man on the footbridge “ and those already
at risk in a situation “ say, those tied to the tracks in the same area.
Unger thinks that this is a cognitive distortion.15 But Hanna thinks
that it is appropriate.16 MAC suggests a kind of middle course. Cer-
tain groups “ namely, those constituting group agents or recognized as

14 Harris (1975) introduced the possibility of alternative risk-sharing practices.
15 Unger (1996: 96“101).
16 Hanna (1992: 320“336).

salient by speci¬c group practices constituting group acts “ are quite nor-
matively salient, and may in different cases contribute different sorts of
normatively relevant factors. But other intuitive groups may introduce
cognitive distortions.
By way of contrast, consider Hanna™s general account: There are moral
situations to which some people are bystanders and in which some people
are participants. It is impermissible for anyone to force a bystander to
become a participant victim in a trolley case, but it is permissible for
a participant to harm other participants in order to establish a better
outcome.17 The fat man on the footbridge is a bystander, but anyone
already tied to the relevant tracks is a participant. So it is okay to turn the
trolley and save ¬ve at the cost of one, but not to push the fat man onto
the tracks.
I think that this is more or less right. Hanna™s account of moral sit-
uations is expressed in a way that suggests rich, and I think implausible,
metaphysical presumptions.18 But I believe this to be an avoidable aspect
of his presentation of the view. Still, while he provides hints about the
distinction between bystanders and participants,19 he does not provide
an account of that distinction that would allow some independent ratio-
nale for our intuitive normative judgments about the cases. He instead
suggests that we should work backward from our intuitive normative
judgments about cases to the distinction.20 This is problematic, because
sometimes our intuitive judgments are affected by cognitive distortions.
Remember the importance of the order of presentation of cases, and the
other forms of incoherence in our judgments about cases noted earlier,
which sometimes seem to be due to distortions introduced by grouping.
What™s more, we need an account that rationalizes our probative intuitive
judgments about such cases, not an account that merely summarizes our
intuitive judgments.
MAC provides a criterion for the distinction between properly rele-
vant participants and bystanders that yields a rationale for our probative
intuitive reactions to cases, and which suggests that while certain intuitive
groupings introduce cognitive distortions in the way that Unger claims,
many do not. It provides a kind of re¬nement of Hanna™s general pro-
posal, but one that also invokes speci¬c distinct factors of the sorts noted

17 Ibid., 330.
18 Ibid., 329.
19 Ibid., 329“330.
20 Ibid., 333.

earlier, and that provides a reductive criterion for the bystander-participant
distinction yielding a suitable rationale for a coherent and properly intu-
itive response to all trolley cases.
Still, so far I have only made rough suggestions about a kind of mecha-
nism that MAC can apply to trolley cases. What may be more telling is an
application of MAC to deliver some particular commonsense intuitions
about differences between particular cases. As I™ve said, unlike standard
attempts at uni¬ed explanations of all our intuitions in trolley cases, MAC
suggests that a variety of small factors that involve group action in different
speci¬c ways can work together to generate signi¬cant differences in the
moral status of different actions. But we need to see how this works in
more detail.
Let™s begin with cases that are intuitively far apart. Foot™s original Trol-
ley case and Transplant certainly occasion different intuitions. But notice
that a trolley driver is a professional who is plausibly specially respon-
sible for the trajectory of the trolley by virtue of a speci¬c cooperative
practice. This may constitute a group action, underwriting a special pro-
fessional obligation to turn the trolley in the direction in which it will do
least harm. If that isn™t evident, we will discuss the interaction of MAC
with special obligations in section IV. Notice also, coming from the other
direction, that the doctor in Transplant is also a professional member of
a cooperative practice providing health care “ indeed, of a number of
overlapping cooperative practices of that sort constituting various group
agents “ who has special obligations to patients in the hospital, and in
general not to cause death.
But those are not the only relevant moral differences between these
two cases. To the end of identifying some of the other factors, consider
intuitively intervening cases “ for instance, Bystander and Fat Man. It
seems morally permissible, or even perhaps required, to turn the trolley in
Bystander, but morally impermissible to push the fat man onto the tracks.
How can MAC deliver this difference?
We have, I think, a real intuition that supports the Pauline Principle,
the principle that you should not do evil (like pushing the fat man in front
of the trolley to his certain death) even if a greater good would come of it
(the life of the ¬ve), while on the other hand it is sometimes appropriate
to do some great good while foreseeing that a smaller evil will come of
it (like turning the trolley in Bystander). We also have, I believe, a real
intuition that you shouldn™t drag an innocent person like the fat man
into a situation like the one in question against their will. We need to
be a little careful, however, in how we handle these intuitions. For one

thing, in real-life situations, when you do evil as a means to the good, the
temporal priority of the evil makes it more certain than it would be if the
temporal order of the good and evil were reversed. It is hard to control
for the effects of this fact when considering hypothetical cases. Second,
I have already noted that not all our intuitions about cases like these can
be treated as probative. Still, there is something behind our intuitions
that suggest some normative weight for the Pauline Principle and for the
distinction between doing evil and refraining from doing good. But MAC
can deliver that weight.
Recall, in this order, Fat Man, Sidetrack, Loop, and Bystander. We
start with a case in which action to save the ¬ve seems intuitively quite
wrong, and then end up with a case in which it seems permissible and
even maybe required. Here is an analysis of this phenomenon according
to MAC:
Between Fat Man and Sidetrack, these factors are relevant: (a) There
is a basic scheme of group action of which almost all of us are partici-
pants, which is a kind of cooperative agreement not to draw uninvolved
outsiders into such a situation against their will. If our generally accepted
background practice were different “ if, for instance, we had all agreed
to share risks in some more rational but still intuitively very scary way “
that would make this factor irrelevant. But under actual conditions it is
relevant. And it is apparently very weighty by the basic normative tenets
of MAC. (b) There may also be a second set of factors that MAC can
deliver in at least some cases that meet the general description we are
considering, though I am less con¬dent of this. The fat man in Fat Man
is outside the originally affected group. But at least if the people on the
tracks are conscious, they may share acceptance of reasons supporting
group action to get at least some of them to safety. They may be a group
agent of one kind. Of course, on the other hand, the practical salience of
this project may be minimized if they are tied down, and a person tied
to a sidetrack will only plausibly share a relatively abstract group project
and accepted reason with those on the main track. But there may also be
another intuitively salient group that includes those who are stuck on the
track and the choosing agent. And it is plausible that the fat man will not
share acceptance of the relevant reasons with that group. He would then
be in a recognizable sense outside of the relevant group that includes all of
those on the tracks, and the other relevant group that includes those on
the tracks and the choosing agent. And these groups correspond to cer-
tain group agents in my sense, which it is reasonable to think have proper
projects. Notice how the normative details according to MAC will turn

on subtleties about the reasons accepted by the various individuals present
in such a case, as well as other psychological conditions, their attempts,
and the normative weight of their group projects. But these complexities
do not seem obviously unintuitive.
These two factors, or at least the ¬rst, introduce some weak normative
differences according to MAC. But weak differences are the most that
commonsense intuition ¬nds between these cases anyway. As we move
from Sidetrack to Loop, all those on the track seem even more closely
linked by these weak factors.
Between Loop and Bystander, there are various factors that interfere
with the probity of our intuition “ for instance, the unreality of the very
clear risks in trolley cases and also temporal prejudice. But at least the
temporal factors might introduce a genuine difference accommodated by
MAC, since almost all of us plausibly share a general scheme of group
cooperation in accord with something like the Pauline Principle that
plausibly constitutes a form of group action. We have a practice in which
it is one thing to de¬‚ect a harm so that it kills someone else and quite
a different thing to de¬‚ect the harm by killing that someone. And that
practice is a normatively weighty group act we share, with a plausibly
proper project.
I have been exploiting intuitive advantages that accrue to MAC because
of our standing cooperative practices for distributing risks and harms.
Alternatives are scary and unfamiliar. Still, these actual practices have
costs also, and possible alternative practices might conceivably have bet-
ter consequences. It is relevant according to MAC which cooperative
practices of this sort we actually have, as long as they are group agents,
but they must also have proper projects, must have basically bene¬cent
results, in order to carry normative weight according to MAC. And it
is also relevant that there may be prospective and alternative forms of
cooperation that would have better results. Of course, our practices for
distributing risks and harms are plausibly one-off group agents. And then
the question is merely whether they are better than their absence. But
even so, I believe that the background practices that give priority to
leaving the fat man undisturbed are at least of questionable propriety
according to MAC when it is allied with some plausible basic conse-
quentialist principles, including HMP. But if they are properly bene¬-
cent “ which, after all, common sense does presume “ then MAC can
deliver the intuitive differences in our reactions to Fat Man and Bystander.
And if they are improper, then commonsense intuitions on this topic are

There is no deep difference between killing and letting die introduced
by the mechanism of MAC, but rather a series of small differences that are
indeed dependent in certain ways on our contingent practices and their
results. We have seen a variety of small phenomena that intervene between
Transplant and Bombing and between the other four cases that are in
some intuitive sense between them. They create normative differences
according to MAC, which track intuitive differences between the cases.
There are other, similar factors that MAC suggests may be relevant in
other cases. For instance, we intuit that the presence of other human agents
in a causal chain breaks to a degree our responsibility for the outcome of
that chain, at least if they are not turned into mere instruments or means
by threats or the like. But MAC might cede those other human agents
a special normative role of that sort if there is a form of general group
agency that makes it especially those others™ business to interrupt such a
chain. This might be assured by a general cooperative scheme, analogous
to those for distributing risks and harms, which is a group agent with a
proper project, or by the structure of a more particular and local group
agent. To the degree that someone in the intervening chain fails to be
a relevantly cooperating agent, this might undercut, according to MAC,
the signi¬cance of their presence in the chain for the commutation of our
responsibility for the relevant outcome, though that would depend on the
details of the project of that group agent.


We have seen that MAC can deliver agency goods like true belief, and
that it underwrites to some degree intuitive commonsense distinctions
between killing and letting die and the like. But there is one last and key
aspect of the standard deontological objection to act consequentialism
under consideration. I will focus in this section on this aspect, on the
agent-centered, deontological restrictions themselves. And I will treat in
some detail the particular cases of lying and murdering.
MAC delivers agent-centered restrictions that are yet supported by
direct consequentialist evaluation of options. Begin with this case: Nearly
all of us participate in a group act of refraining from torture, which has
signi¬cant normative weight even according to a classical utilitarian val-
uation principle or HMP. So VLD and DD will seldom countenance
defections from that group act when conjoined with such a basic norma-
tive principle, even if a particular act of torture would serve the good “
indeed, even if it would prevent several other tortures.

And this point can be extended. Group acts of refraining from violation
of various intuitive deontic duties that directly protect agency goods like
health and knowledge will yet characteristically have weighty overall pos-
itive consequences even according to plausible forms of hedonism. And
remember that for one-off group acts this is all that is required to invoke
the protection of VLD and DD. And we saw in section I that MAC gen-
erates various other kinds of respect for “genuine” agency goods, which
support proper projects.
Crucial instances of agency goods are specially constituted or protected
by familiar deontological restrictions on action that are delivered by the
mechanism of MAC in these general ways. Some of the negative restric-
tions that we previously noted as crucially helping to constitute and foster
these agency goods are quite often at least very close to full-blown agent-
centered restrictions. Violation of standard deontological prescriptions
is in fact often defection from standing group acts with quite weighty
projects that we are almost all inside, and hence often a violation of DD
or VLD. In actual fact, we accept agent-constituting and indeed higher-
order agent-governing reasons of signi¬cant consequentialist weight, even
as assessed by plausible hedonistic valuation principles, that re¬‚ect deon-
tological restrictions, and MAC will endorse such agent-constituting and
agent-governing reasons that enjoin deontological restrictions on action
when one is inside groups with proper projects unless violations are sanc-
tioned by DD or VLD. While this requires that the groups in question
have appropriate projects, there are many forms of one-off group action
that are forms of moral cooperation with good effects, the cooperative
creation of spheres in which the standard deontological prescriptions are
properly observed. Some are informal and local “ say, when you and I
have a special practice of not lying about certain things to each other. This
may be local enough to constitute a special obligation, and we will turn
to such cases in the next section. But some are almost universal, regarding
both those who cooperate and those to whom the relevant respect is due
according to the project of the group. They re¬‚ect contingent but rela-
tively broad social conventions. These help constitute general obligations
that are yet dependent on the existence of speci¬c group acts. These are
the key group acts for handling deontological objections to MAC.
What if you are outside the relevant groups? What if you are a congen-
ital liar? I think that in fact almost no one is outside the relevant groups,
because almost everyone accepts the relevant reasons to one limited degree
or another. But, as I noted in the last chapter, MAC suggests not only that

you should cooperate with the weighty group agents that you are inside,
but also that you should join the weighty ones that you aren™t inside
if that is suggested by the true consequentialist valuation principle “ in
other words, if you can provide more additional value by joining than you
might alternatively produce “ as long as that doesn™t violate other strictures
of MAC. Once you are inside, you will then be tied in with something
like deontological force. That is what you should do if you are outside
the relevant groups, but what are you owed? Though there are no doubt
differences in the treatment of liars and nonlying cooperators enjoined
by our standing group act of refraining from lies, still you deserve at least
some truthful dealing.
Let me treat a speci¬c case in at least a little detail to show how this
mechanism can work. In particular, let™s continue with the case of lying.
There are four key facts: (i) We are inside a variety of small-scale group
agents with properly bene¬cent projects and such that to lie within the
group is in fact to defect from the group project, as well as other small-scale
agents in which it is somewhat indeterminate or vague whether an internal
lie involves a defection. (ii) The proper projects of other bene¬cent agents
should be supported by truth telling. (iii) But most important is that almost
everyone is part of a very weighty group agent that is a one-off practice
forbidding lies, except under various somewhat vaguely speci¬ed excusing
conditions. And defection from that practice is often all too determinate.
(iv) What™s more, those few outside the practice will generally have a
reason to join, on straightforward act consequentialist grounds. And if
they join, then they will be bound inside.
There are subtleties that attend the last point: It is not, I think, that
an atomic agent that is some reformed moment in the life of a former
liar is speci¬cally bound by the reforming choices of their past selves, but
rather that to join a group agent is to come to accept relevant reasons,
and in a reformed moment the agent ¬nds itself inside the relevant group,
by the natural temporal inertia of accepted reasons. And of course if
some pill for changing the reasons we accept becomes available, it will
still characteristically constitute defection from a group act to try to put
your future selves outside it. And, as I noted in the last chapter, it is not
merely atomic agents who join moral practices. But let™s concentrate on
the crucial normative implications of my four key points.
These four general considerations suggest the following four relatively
concrete tenets as directives of MAC, which in the context of our actual
practices are quite signi¬cant: (1) Do not lie, and more generally, do not

create false expections or fail to deliver relevant information, within a
group agent with a properly bene¬cent goal that lies outside of the prac-
tice of cooperation itself,21 unless a violation of that duty is sanctioned by
VLD or DD, or the lie serves the project of the group, or the lie is accept-
able in the practice. (2) If other things are equal, do not lie, and more gen-
erally, do not create false expectations or fail to deliver relevant informa-
tion, to another agent or group agent with a properly bene¬cent project.
(3) Do not lie, and more generally, do not create false expectations or fail
to deliver relevant information, whenever you are within a one-off group
action creating a sphere of truth telling of some sort that is itself a proper
project, unless a violation of that duty is sanctioned by VLD or DD, or
unless the lie is acceptable in that practice. (4) Join such groups whenever
that is required on direct consequentialist grounds, except when that is
forbidden by MAC.
Focus on the third directive. You and I are in fact within such a
widespread one-off group agent, which is quite weighty according to
most plausible consequentialist normative principles, and so VLD and
DD will very seldom sanction exceptions from that practice. The practice
itself, of course, speci¬es some appropriate exceptions from rigorous truth
telling, though often only vaguely.
It is normatively quite signi¬cant according to MAC that we have a
general cooperative practice that is a group agent generally forbidding lies
though with various exceptions, and that does allow a certain amount of
misleading speech. Roughly speaking, you should lie in a way forbidden
by that practice only when you can gain as much value by that defection
from our general group act as that whole vast cooperative practice creates.
Still, there are possibly more weighty practices in which one could
engage in a way that contributes value by more strenuous truth telling, and
strenuous truth telling will very rarely require defection from a less strenu-
ous cooperative practice of truth telling. Independent of the actual details
of our group practices and the contingencies of consequences, the differ-
ence between explicit and implicit lies and the difference between lying
and withholding information are irrelevant according to MAC. MAC
suggests that bene¬cent projects of any agent should characteristically be
supported by truth telling.
Yet there are limits to the kinds of strenuous truth telling required by
MAC. Perhaps it will seem that MAC cannot endorse enough lies. We do

21 This phrase distinguishes this obligation from that expressed by (3).

not plausibly gain more by lying to a murderer searching for an innocent
victim than an entire practice of not lying gains the world. But still,
our actual practice about lying endorses such exceptions, and indeed has
better consequences than alternatives that are more rigid in the notorious
Kantian way. So go ahead and lie to the murderer.
There is another sort of balance between lying and murder that we
should consider, and along the way we can treat a second relatively con-
crete intuitive deontological restriction. It will take us a moment to get
back to the issue of balance.
A precept forbidding killing is found in all standard intuitionist accounts
of commonsense morality, but there are differences about killing animals.
Traditional consequentialist hedonism in conjunction with MAC suggests
that animal pain is normatively signi¬cant, but that killing nonhuman ani-
mals isn™t especially important except in the unusual case in which it vio-
lates some form of mutual cooperation that is group action, which would
be possible with other animals capable of speech and hence of accepting
reasons in my sense, or in case it is speci¬cally prohibited by bene¬-
cent elements of the project of a bene¬cent group act. Killing humans
is another matter. Killing others with whom one is engaged in group
activity of course eliminates the possibility of cooperative activity. Even
in the unlikely event that the project of the group is favored by the murder
of a single member, still this involves a violation of cooperative activity
in pursuit of that goal, unless that possibility is speci¬cally endorsed by
the whole group and in particular the member murdered. And of course
there is a general cooperative practice among humans of not murdering
one another, which is a one-off group agent with a very weighty proper
project, though it seems to allow “ or at least it seems not determinately
not to allow “ certain exceptions for wartime, self-defense, and the like.
It is also relevant that participation in this general form of cooperation
is a likely precondition for admission into most forms of other coopera-
tive activity, and that killing others will undercut their individual proper
projects and those of the group agents of which they are part.
With those factors noted, we can see that the following analogues
of our earlier precepts governing lying will be endorsed by MAC as
weighty reasons: (1) Do not kill human or nonhuman cooperators within
a group agent with any properly bene¬cent project having a goal that lies
outside of that practice, unless a violation of that duty is sanctioned by
VLD or DD. (2) If other things are equal, do not kill humans or other
agents who have (or are parts of group agents that have) proper projects.
(3) Do not kill whenever you are within a one-off group agent creating

a sphere of nonmurder that is a proper project, unless a violation of that
duty is sanctioned by VLD or DD, or is sanctioned by that practice as an
appropriate exception. (4) Join such groups whenever that is required on
direct consequentialist grounds, unless that is forbidden by MAC.
You and I are within such a practice as the third precept invokes.
Since we have a general cooperative practice forbidding the murder of
humans under most conditions, and which is a group act that is quite
normatively signi¬cant by the tenets of MAC, MAC largely forbids such
murder outright. You may murder to save the entire world from certain
destruction, though probably you cannot murder an innocent even to
stop a small war.
A characteristic sore spot for MAC is that when normative practices
that are group acts con¬‚ict for an agent, it may be that the practice that has
the weightiest consequences is a practice from which individual defection
will be less weighty. And even individual defection may often achieve
weighty goods. For instance, it may stop a war. Other possible examples
of this general sort involve the interaction of lying and murder. And so
we return to our question of balance.
You may of course lie to save a life, because that exception is clearly
built into our truth-telling practice. And a more rigorous practice of
truth telling would not have better consequences. Surely our practice also
allows you to lie to avoid committing a murder, if there can be any such
case. But if we were inside a group agent with a proper project that did
not allow for these exceptions, then MAC also would not allow them,
unless through defection you could help to create a weightier group act
that incorporates the exceptions, or unless the group act of refraining
from murder is weightier than the group act of refraining from lies. I
believe that it is weightier in fact, so that there are two sorts of pro-
tection from murder by truth telling according to MAC, even beyond
the unreality of a case in which our forced choice is to lie or to mur-
der. But imagine that there are in fact situations in which the choice
is between lying and murdering and must involve defection from some
deontic practice, and that while to defect into murder individually is
very weighty in a negative way, still the general group act of not lying is
weightier than the general group act of not murdering, perhaps because
we are not much inclined to murder even independent of that moral
practice. Then MAC would tell you to murder rather than to lie. And,
of course, that would be wrong. MAC characteristically decides between
con¬‚icting deontic duties by appeal to DD, by ceding priority to the duty
that is part of the consequentially weightiest group act; and while that is

at least generally intuitive, it is reasonable to worry that there may be
Or consider again this abstract possibility: Perhaps you are part of two
con¬‚icting group acts that in particular constitute two deontic practices,
and perhaps the ¬rst is somewhat weightier but would survive your defec-
tion, while the second would not. Yet DD prescribes that you stick with
the ¬rst. That seems to me wrong also.
Still, I know no realistic and intuitively debilitating examples of these
two sorts involving the con¬‚ict of two general deontic obligations, in part
because intuitive deontological restrictions are characteristically negative,
and in part because weighty normative practices would characteristically
survive individual defection. Nor would participation in a weighty but still
not nearly universal bene¬cent group act “ say, a political party in pursuit
of world peace “ trump such deontic restrictions according to MAC,
unless it could (implausibly) generate more weighty results than the entire
deontic practice in question. And remember what I said in the last chapter
about implicit exceptions and the Big and Small Disasters. Nevertheless,
we will return again to dif¬cult cases with this general structure during
our discussion of special obligations.


We have seen that MAC can evade one standard objection to act con-
sequentialism. But a second standard objection to act consequentialism
is its apparent failure, in Thomas Nagel™s words, to respect the special
obligations “we have toward those to whom we are closely related: par-
ents, children, spouses, siblings, fellow members of a community or even
a nation. Most people would acknowledge a noncontractual obligation
to show special concern for some of these others “ though there would
be disagreement about the strength of the reasons and the width of the
net.”22 There are also intuitive special obligations due to our speci¬c past
acts “ for instance, duties to keep promises and to obey contracts, and
duties of reparation and gratitude.
There are differences of opinion about the force and nature of these
obligations, and indeed some internal confusion within most individu-
als about them. How much should we favor those to whom we intu-
itively owe gratitude? And there are a number of quite relevant traditional

22 Nagel (1986: 165).

consequentialist observations, deployed in traditional consequentialist
responses to this objection. For instance, we have special ability to favor
the interests of those we know well and to whom special affection ties
us, as indeed to harm them by neglect. And of course consequentialists
should consider the actual background for any action, in which in fact
most act mostly to favor their own.
But Multiple-Act Consequentialism suggests the relevance of another
class of factors that are very helpful in response to this second objection.
Many intuitive special obligations parallel plausibly proper reasons accord-
ing to MAC. There are different ways in which this is so. But the main
point is that special obligations are often within relatively small group
agents, and often help to constitute effective and appropriately valuable
group agents. In fact, some of the explanations for intuitively general
agent-centered restrictions that we surveyed in the last section are also
explanations of special obligations dependent on the existence of par-
ticular, though sometimes very large, group agents. We probably have a
somewhat special obligation not to murder other cooperating nonmur-
derers according to MAC, though that depends on the details of our group
project of refraining from murder. Narrower and more intuitively special
obligations are created by participation in smaller bene¬cent group agents
in exactly the same way.
Acceptance of particular special obligations may be acceptance of
agent-constituting reasons that help to constitute the existence of par-
ticular group agents “ for instance, a particular cooperating family, or a
group act of two bound by a promise and by whatever reciprocal condi-
tions elicited the promise. Of course, the normative details turn, according
to MAC, on details of the psychologies of the parties involved, as well
as on the normative weight of these group agents. So no special group
agent is created when a lying promise is made. But the general deontic
practice of refraining from false promises and other lies is of course itself
a very weighty group agent in which nearly everyone takes part and that
others have direct consequentialist reason to join.
Acceptance of special obligations may also be, more narrowly, accep-
tance of agent-constituting reasons that help give a family a certain appro-
priate project. Or, more broadly, it may be the general acceptance of
agent-governing reasons of the appropriate sort, at least given plausible
contingent but general conditions that imply our special ability to harm
or help those to whom we are closely tied. Given the general recogni-
tion of a general obligation to support effective social practices of child
care, those contingent conditions may imply the need to accept a special

obligation. And acceptance of general forms of special obligation “ say,
general obligations of gratitude “ may constitute someone as an effective
candidate for many forms of group action. What™s more, standard forms of
social arrangement often involve generalized group action that supports
particular sorts of smaller group agents, or that singles out the normative
relevance of certain factors to their proper functioning.
One quite characteristic and at least generally intuitive feature of MAC™s
treatment of special obligations follows from the fact that small group
agents “ say, particular friendships “ ordinarily have less weighty conse-
quences than massive group agents, such as the general practice of refrain-
ing from murder. So, by the principle DD, MAC will seldom allow special
obligations to trump the deontological restrictions we discussed in the last
But this structural feature of MAC may seem to raise an intuitive dif-
¬culty in some cases that are analogous to the con¬‚ict cases we discussed
at the end of the last section and in the last chapter, but in which the
con¬‚ict of special obligations and weighty bene¬cent group practices is
at stake. Imagine that someone™s role in a group agent requires that he
or she work on Wednesday afternoons toward a charitable goal, and that
the group agent does massive amounts of good, but that any Wednesday™s
contribution from that individual is relatively trivial. Now imagine that
on a Wednesday there is a choice for that individual between saving their
own son from drowning and performing their role in the charity. It may
seem that MAC requires that the parent let the son drown, and that of
course would be wrong. But, as I noted, special obligations according to
MAC involve not only small-scale group agents that constitute particu-
lar special groups, particular families, but also larger and hence plausibly
weightier group agents that are social practices supporting and govern-
ing such particular group agents. And our very widespread practices of
that sort are weightily bene¬cent group agents that mandate an intuitive
treatment of the con¬‚ict in question, and that plausibly trump the effects
of any speci¬c charity. And there is also plausibly a general obligation
to rescue in emergencies that re¬‚ects a group agent that also trumps the
weight of the charity according to MAC. What™s more, recall that in cases
where we have no tendency to criticize defections from a group act in
favor of a pressing alternative, there is an implicit exception within the
project of that group act.
Despite these various analytical complexities, special obligations, on
the conception suggested here, are primarily proper agent-constituting
reasons that create certain forms of relatively local group agents.

Consider a social arrangement with many families. Each family may
be a group agent. In a typical case, it may be a family crucially and
centrally engaged in a project of mutual care extending especially to the
young and the old. When engaging in that project involves the acceptance
of supporting reasons, which seems the common case, they are agent-
constituting reasons. Of course, the family may also have other group
projects, which involve acceptance of project-constituting reasons. Or it
may be that even in the absence of such a locally bene¬cent central project,
some family still cedes special normative authority to the decisions of some
elder; it accepts them as reasons, as agent-constituting reasons. Either
family suggested is a group agent. And a supporting social arrangement
may itself be the group action of a larger group agent, cooperation by all
on a form of child raising.
Partly because of the nature of standard background arrangements,
which require that families care for their own children and elderly mem-
bers in order to assure good care, partly because of the natural effects
of propinquity and natural sentiments on the effectiveness of this general
strategy, but mostly because the care of the young and the old is an impor-
tant project by the direct light of standard consequentialist basic principles
and because the effective creation and education of moral agents is a crucial
moral task, the central project of the ¬rst family noted in the last paragraph
is characteristically in proper accord with standard basic consequentialist
principles. The project of that ¬rst family is a proper project.
It may seem that we have special obligations to help family members
even in activities that aren™t essential to the family or part of any of its
group projects “ say, to help them pursue a sport on their own that has
nothing to do with any project of the family, or to help them make a good
career move to another city. And such obligations may seem unavailable to
MAC. But notice that these obligations are often folded into the general
standard central project of caring for and nurturing the young and the old,
and that when they are not so folded, they seem intuitively less pressing.
MAC also suggests characteristic complications involving proper
projects. If the central project of a family, perhaps because of its polit-
ical or economic power, or because of its cruelty to its members, is not
in general accord with the basic consequentialist normative principle,
then MAC would suggest that the special obligations in question would
be attenuated or overridden. There might be agent-constituting reasons
accepted that make such a family exist as a group agent, but they would
be improper. And the initial asymmetry in power of parents and chil-
dren suggests that vicious parents can easily assure this. Still, it may also

be that each individual family has a central project that plays a role in
some higher-order group agent, a project for each family that is not itself
in direct accord with the basic consequentialist normative principle, but
such that the project of the higher-order group agent is proper. And there
may be several different group agents with different central projects that
overlap in the same intuitive family, and some of them may be proper
while others are not.
As we have already noted for some dif¬cult cases, a second characteristic
complication of MAC is that it suggests that we must properly balance this
form of group agency, even when it has a proper project, against others.
Indeed, there is more than one form of group agency automatically in
play even in the case of a family with a proper project. There is a particular
cooperative project of some family, and then the existence of that group
agent, and then the larger group agent cooperating on a social form.
MAC balances forms of agency by favoring the normatively weightiest
form, and hence will in some instances “ say, where the very existence
of a family is more signi¬cant than some small failing even in its central
project “ cede that greater signi¬cance. Need it characteristically cede
greater dominance to the larger social project, the cooperative form of
families? That is in one sense a characteristic suggestion of MAC, but
the issue introduces many complexities. There may be many competing
forms of possible family structure, with some more entrenched but others
more bene¬cent. And the kinds of group agents that MAC deploys are
timeless, so that relevant cooperative practices can be prospective. And
the actual form of a family may put it outside of some larger social form
without involving a con¬‚ict within atomic agents of the sort that engages
the characteristic tenets of MAC. And of course the larger social form
may not, on balance, represent a group agent with a proper project, or
even any sort of group agent. MAC™s analysis of special obligations is
complicated in what seems to me a suggestive and properly intuitive way,
and provides a plausible framework for further normative exploration.
MAC naturally treats many familiar institutions as group agents, which
hence invoke this general analysis of allied special obligations. Indeed,
it may be worth noting that there is a natural analysis of all institutions
that is suggested by MAC, one that itself suggests that they are often
group agents in my sense.23 Let me approach it by way of a contrast.
Rawls suggested that an institution is a “public system of rules which

23 This is developed at greater length in Mendola (1988).

de¬nes of¬ces and positions . . . [and speci¬es] certain forms of action as
permissible, others as forbidden. . . . An institution exists at a certain time
and place when the actions speci¬ed by it are regularly carried out in
accordance with a public understanding that the system of rules de¬ning
the institution is to be followed.”24 But there are two problems with this
account as it stands, which seem to demand some generalization. First, it
concerns only systems of rules that de¬ne speci¬c of¬ces and positions,
such as judge, senator, batter, or umpire. But not all social communities
have such a form. Consider a group of agents who accept rules about
general courtesy. We can imagine less institutionalized forms of society
than our own, where there are many fewer rules about the privileges and
duties of particular of¬ces, less division of labor. Second, Rawls™s rules
specify certain forms of actions as permissible and forbidden, rather than,
say, as better or worse. They are supported by penalties and defenses rather
than, say, by a sense of excellence. But we can certainly imagine forms
of society in which a conception of what is good, as opposed to what
is permissible or wrong, plays a dominant organizing role. If we stretch
Rawls™s notion of an institution to include shared conceptions of the good
as well as of the right, and if we don™t require a division of labor, we should
perhaps cease to speak of rules. The general phenomenon of which rule
following and talk of the good are parts seems to be normative evaluation,
citing and accepting reasons for and against things. So here is my proposal:
A social community, any social community, is a group of agents who
accept the same reasons. It is not that agents must share acceptance of
all reasons in order to form a community, but that there corresponds, to
any set of reasons accepted by an intuitive group, a particular community.
This is a weaker condition than is required for group action, but it would
seem that in humans accepted reasons naturally assume that other, richer
form. The accepted reasons naturally engage our sentiments of reciprocity
and benevolence. So many communities are not just groups of agents,
but also group agents in the sense speci¬ed in the last chapter, and they
are of greater normative weight according to MAC than other forms
of community, at least when those other forms of community are not
recognized as signi¬cant in the content of a group act.
The general pattern of analysis of special obligations I have suggested
here also provides MAC with an intuitive conception of proper punish-
ment, which in turn underlies its conception of obligations of gratitude

24 Rawls (1971: 55).

and reparation. We are enjoined by MAC to cooperate in a variety of
normatively weighty one-off schemes like that of not murdering other
humans. But if someone violates such a cooperative practice, MAC implies
a change in their normative status. Depending on the existence of other
weighty group actions in which the perpetrator continues to engage, and
depending on whether the one-off group agent of refraining from murder
involves a directive to refrain from the murder of even noncooperators,
the propriety of the individual punishment of perpetrators may then even
turn directly on the consequences of such punishment as evaluated by the
basic consequentialist principle. In effect, under realistic conditions, given
the kinds of general group acts in which we engage, MAC implies a kind
of automatic normative punishment for those who violate certain forms of
normatively weighty cooperation that are group acts, which is a change
in normative status. This may in turn defeat a normative presumption
against a more concrete form of punishment.
This sort of normative punishment applies also in less weighty cases
than those that would properly occasion signi¬cant institutional punish-
ment such as imprisonment. And this in turn suggests a natural analysis of
duties of reparation and gratitude. Duties of reparation are at least some-
times duties to restore forms of group action that one has violated, or,
more speci¬cally, to make it be that in fact one hasn™t in the end really
violated a reciprocal obligation in the manner that one™s past activity alone
would suggest. Duties of reparation can be in this way a present shadow
cast by the combination of one™s former cooperative obligations within
a group agent and one™s past action in apparent violation of those obli-
gations. And duties of gratitude may sometimes be duties to avoid the
failures that occasion duties of reparation, the shadows of shadows, as
well as straightforward cooperative obligations that root later duties of
There are many details that require further exploration. But special
obligations and analogous phenomena do not provide an evident dif¬culty
for MAC, no more than deontological restrictions. They just involve
smaller group agents.


The ¬nal objection we must consider is that act consequentialism seems
too demanding. Indeed, it may even seem so demanding as to undermine

individual integrity. Bernard Williams persuasively reminded us that a
is identi¬ed with his actions as ¬‚owing from projects and attitudes which in some
cases he takes seriously at the deepest level, as what his life is about. . . . It is absurd
to demand . . . that he should just step aside from his own project and decision and
acknowledge the decision which utilitarian calculation requires. It is to alienate
him in a real sense from his actions and the source of his action in his own
convictions. . . . It is thus, in the most literal sense, an attack on his integrity.25

But there are two crucial things to notice about this class of objections.
First, it is a class, and has different forms with more or less unequivocal
support from re¬‚ective common sense. Second, the unequivocal forms
can be evaded by Multiple-Act Consequentialism. If consequentialism
literally undermined one™s individual integrity as an agent, that would
indeed be deeply counterintuitive. But if consequentialism just erects a
few barriers to the unbridled pursuit of the most trivial sorts of self-
interest, that is a different matter. If a moral theory implies that we must
spend a bit more on famine relief than the ordinary person spends, when
that has a trivial effect on their self-interest and well-being, then we do
not immediately know on intuitive grounds that that moral theory is
hence false. Any intuition to the contrary is so obviously self-serving as
to be immediately suspect. Still, there are more intense demands that a
moral theory might make, and that are intuitively disturbing demands of
some forms of consequentialism. But MAC does not underwrite those
First case: Familiar act consequentialism may allow or require that in
pursuit of famine relief we routinely violate intuitive special obligations
to our children and general deontological obligations not to steal or to
deceive others. But as we saw earlier, MAC delivers intuitive special obli-
gations and deontological constraints, though with various characteristic
Second case: It may well be that familiar act consequentialism requires
so much of us that it is sometimes in the most literal sense an attack on
our integrity, as Williams suggested. But Multiple-Act Consequentialism
does not require that much.
There is some ambiguity regarding what Williams meant, and we
should consider more than one reading. We might read him as com-
plaining that consequentialism literally demands of act consequentialists

25 Williams (1973: 116“117).

the cessation of their own moral agency. But anything literally required
by one™s continuing moral agency, or indeed by one™s continuing effective
moral agency, at least when that agency is in the service of appropriately
bene¬cent projects, is characteristically protected by MAC.
According to MAC, one must characteristically treat one™s future selves
with suf¬cient respect so as to constitute oneself as an effective coopera-
tive group agent over time. Agent-constituting reasons that help root that
uni¬ed continuing agency often trump direct consequentialist action by
an atomic agent. Of course, there are complexities that attend the bal-
ancing of agents required to underwrite respect for one™s future selves.
First, there seems to be a weighty group agent in which we nearly all
participate that enjoins at least minimal concern for one™s future selves,
whatever their projects. But second, if one™s obligations to one™s future
selves are not rooted merely in such a general group agent, such addi-
tional respect characteristically depends on one™s acceptance of a tem-
porally persisting, effective, and proper project. It also depends on the
demands of proper agent-balancing reasons. Sometimes the project of
the moment may be more normatively weighty than a long-term con-
sequentialist project. There may even be times when one™s continuing
agency must properly be sacri¬ced. If one reliably pursues bad conse-
quences, that may often be so. But we have no intuitive consensus that
there aren™t any times when morality is demanding in this way. What-
ever Kant would say, it is not unintuitive that morality demands that the
autonomy of the immoral sometimes be undercut. If I am so deeply and
ineluctably immoral that I cannot at will become a properly moral agent,
then I should do what is necessary to tie myself to the mast. Despite these
various complexities, as I said, MAC characteristically requires that one
treat one™s future selves with the respect due to cooperating agents sharing
a bene¬cent group project. MAC requires an asymmetrical treatment of
existing group agents of which one is part and merely possible agents. But
existing agents in the relevant timeless sense include cooperators in the
future, and so your future selves.
MAC provides in these ways a general framework to underwrite the
intuitive plausibility of some moral duties to oneself, including one™s
future selves, and also the intuitive signi¬cance of many individual virtues,
including many that root standard virtue-based objections to consequen-
tialism. The two basic points are simple. Such duties to self and virtues can
be enjoined by bene¬cent one-off general group agents, general practices
that meet the strictures of MAC. And one™s own continuing, effective,
and proper moral agency is something that MAC will directly underwrite

with a variety of agent-constituting reasons shared over a continuing
life. This is because one™s continuing agency is a kind of group agent,
constituted in part by these accepted reasons. There are agent-governing
reasons that suggest duties to develop one™s stable and properly effective
moral character. But there are also agent-constituting reasons that suggest
this. And there are other agent-constituting reasons that indeed help con-
stitute, when stably accepted, at least some part of that stable character, at
least some part of the requisite stable dispositions to behavior and feeling
that constitute such a character. Individual virtues are, on the conception
developed here, naturally entwined with a certain crucial form of group
action that you undertake all by yourself. Traditional individual virtues
exhibit the characteristic entwining of reciprocity and benevolence that
we should expect of genuine group action. We want to be fair to ourselves,
and to take care of ourselves, and those motives entwine.
So MAC does not undermine one™s integrity insofar as one™s integrity
is protected by duties to self and often quite self-serving traditional virtues.
But my discussion may seem to ignore what Williams really meant. More
plausibly, he just meant that consequentialism is way too demanding,
that it leaves insuf¬cient room for normal human life outside the main
consequentialist project, even if it can allow room for individual virtues,
duties to self, and general and special deontic obligations. MAC will
endorse a variety of projects of weighty group agents outside the general
consequentialist project of individual good-doing by atomic agents, and
indeed characteristically cedes them greater weight. But that still may seem
to duck the major issue. While MAC won™t characteristically require any
literal sacri¬ce of one™s personhood or of traditional individual virtues and
duties or of all alternative projects, and while it is appropriately consistent
with common sense if MAC is slightly more demanding than our usual
grossly sel¬sh practice, still direct consequentialism may seem way more
demanding than is appropriate.
Just how demanding is MAC? What exactly does it require of the well-
off in a world full of human suffering? What speci¬cally are our obligations
to the starving according to MAC? That depends in part on the general
consequentialist normative principle with which it is conjoined. But it is
possible to say something general.


. 3
( 10)