. 9
( 10)


hadn™t so acted. Sometimes there are single and signi¬cant choices “ say,
social choices of institutional arrangements “ where there are no iterated
probabilities of many actions to worry about, or for that matter to save us
if we have one very bad shake of the practical dice, and in that case worst
possible outcomes seem more intuitively signi¬cant. But still it is likely
that if we act in accord with a risk-averse principle in such a case we will
end up with an outcome that is normatively worse than if we had not been
so risk-averse. Whatever the confusions of common sense on the topic
of normative risk, still HMP may seem to be intuitively unacceptable on
these grounds.
Let me put the same objection in another way. Maximining over risks
is generally rejected in favor of maximizing expected utility. That seems
like paradigmatically rational behavior regarding risks, just as prudently
maximizing one™s total individual utility may seem like a paradigmatically
rational response to the temporal distribution of goods within one™s life.
After sections I, II, and III, we should hesitate in the face of that parallel,
and maximizing expected utility does require a valuation of outcomes that
is not available in our ordinal construction. Still, Ramsey, von Neuman,
Morgenstern, and Binmore have suggested a way to work backward, a
way to construct quantitative individual utility that re¬‚ects what they take
to be intuitively rational preferences over gambles.

But, I reply, this sort of allegedly “rational” expected utility maximizing
is quite a dramatic idealization of real human behavior, just as individual
utility maximizing within a life is an idealization. It is hence open to
suitable normative question and manipulation. And it isn™t in fact rational
in the fully normative sense of that ambiguous word. It matches neither
the way we in fact behave, nor the way we ought to behave. Risk-averse
behavior is constituted as properly rational in the full normative sense by
the argument of the previous chapter. Largely because the basic value that
happens to be found in our world comes ordinally, we can only develop
an ordinal valuation of outcomes, and hence an ordinal valuation of risks
over those outcomes.
Still, it might be countered, agents who act in accord with risk-
averse decision principles are irrational in the different sense that they
will probably lose out according to their own valuation of outcomes,
especially over the long run. But of course at issue here is what is the
nature of the most important way of losing normatively, even in the long
run.41 The individual losers in any forced lottery that promises to max-
imize expected utility overall, even if there is an overall lucky outcome
for that lottery, do not intuitively lose their grounds for complaint just
because expected utility was maximized. And of course the complaints
of the losers if there is an unlucky outcome may intuitively be even
I have made two clusters of points. Common sense is vague on issues
of normative risk, and the theoretical advantages of more traditional con-
ceptions are not decisive, just because it is not obvious that they are the
particular theoretical advantages that must intuitively count most. But
there is also a third set of signi¬cant points.
The most practically relevant implications of HMP regarding norma-
tive risk are mediated by the Proposed Code and hence by Multiple-Act
Consequentialism. And the complexity of this mechanism of application
serves to blunt some intuitive worries about the risk aversiveness of HMP.
Remember that many of the most weighty acts according to MAC are
very large one-off group acts, which are not relevantly iterated. And while
the nature of the Proposed Code will turn on the nature of our options
in various ways, still it will involve a generality that will engage salient
general risks and features of outcomes at the expense of particular details.
Salient and striking general risks do seem to matter to common sense,

41 Here one reason for distrusting self-defeat arguments against normative principles will be
evident. For another, see Mendola (1986).

though not, as I™ve said, in any very precise way. And the salient worst
possible outcomes of many of the relevant general risks properly in ques-
tion in evaluation of the Proposed Code are effectively identically evil.
Then what matters according to our principle is a greater probability of
that worst outcome, which, when it rests on suf¬ciently general grounds,
is likely to be intuitively salient to choice.
To summarize: It is impossible to develop any very determinate account
of normative risk on the basis of unclear commonsense intuition. Stan-
dard theoretical advantages of standard theories are not decisive. And
where commonsense morality has clear commitments on these issues,
they interact with HMP only through the mechanism of Multiple-Act
Consequentialism, which cushions remaining con¬‚ict.


The three controversial structural elements of HMP “ its focus on ¬‚ecks,
its maximin form, and its risk aversion “ interact in intuitively relevant
ways. They also interact with the hedonism of that principle. Some of the
most obvious of these interactions may seem intuitively unfavorable. Let
me put such an objection starkly and summarily.
HMP pays little heed to the way happiness is distributed among distinct
lives, and to the degree that only a single life is at issue, it implies that no
life at all is better in the most basic sense than one containing a moment
of pain. It implies, for that matter, that no sentience at all is better than a
situation in which one creature is momentarily in pain. Indeed, it is better
that there be nothing at all than that there be the slightest chance that a
creature will feel one ¬‚eck of pain. That there be nothing, and certainly
none of us, would be better than any situation that we could plausibly
achieve, let alone any risk we could plausibly run. What™s more, there is
another way, despite all this, in which HMP may seem to give pain too
little heed. It suggests that it is better to have a situation in which everybody
is always pretty miserable than a situation in which one moment, indeed
one ¬‚eck, of one person™s life is worse than that and everybody else is
quite a bit better off.
Despite the fact that congruence with common sense is important,
HMP is not so refuted. I will develop three replies to the contrary con-
First of all, we have already considered some points that blunt elements
of this objection, which we can see by focusing on one sort of pressing and
troubling case. Recall our discussion of Temkin and Scanlon in section IV,

and consider in that light a particular case that may seem quite problematic
for HMP.
Presume that we must rank two worlds. In the ¬rst, there are ten billion
happy people. Not only are they happy on the whole, they are happy all
the time “ except that one of them has an accident, and crushes a ¬nger,
and feels an hour or two of physical pain. In the second world, there is
no sentient being whatsoever. HMP says the second world is better, and
that is not what we intuitively think.
However, remember that our intuition is misled by the lack of respon-
sibility presumed in this case. The properly intuitive normative evaluation
of outcomes is revealed only when we feel full responsibility for those
outcomes. So instead assume that you must torture an innocent by break-
ing his ¬nger against his will in order to bring the other beings into their
happy existence. That is something many of us might do anyway, but it
is not con¬dently commanded by commonsense moral intuition. Deon-
tologists, in particular, may ¬nd it intuitively problematic. Of course, the
innocent whom you torture may thank you next year, just as Ahab may
thank you in the morning for excising his leg against his will. But in
sections I and II we saw that commonsense intuition is also sometimes
misled by a failure to notice the morally relevant difference between a
person at one time and the same person at another time, even though on
full re¬‚ection that difference is of real intuitive weight.
My second reply will focus in part on the kinds of details, presumed
or stated, that make alleged counterexamples that involve the extinction
of sentience, or pain of one for the sake of general pleasure, intuitively
powerful. Hovering in the background of such cases are the possible con-
ditions that would make such choices available to us “ for instance, dooms-
day machines to certainly exterminate all life, or contraptions to increase
someone™s pain just a bit and certainly eliminate much other pain. And
my point is that those details are not located at the level of abstraction
that we currently engage. They cannot be properly discussed until we
consider the interaction of HMP with the mechanism of Multiple-Act
Consequentialism and understand details of the Proposed Code that it
helps support. On these topics, I must now request your patience.
Still, there are some relevant considerations that we can properly see
even now. The concrete conditions presumed in the background of such
objections must be very unrealistic, and in a specially problematic way.
No realistic situation gives us these options. For instance, none of us is
ever going to be in a position to choose an outcome in which there is cer-
tainly no further sentience, as opposed to a wild thermonuclear option

on which there are lots of risks of very painful future sentience. Even
most science ¬ction extrapolations wouldn™t give us that certainty. And it
is also relevant that the certainty is a very precise detail of the cases, but
one crucial to their intuitive force. Ivan™s case is unrealistic, and so too
are the spatial inversion cases we considered in section II. But their unre-
ality doesn™t infect very precise details of the cases that very signi¬cantly
drive our intuition “ for instance, a difference between certainty and near
certainty. Our earlier cases were in that sense appropriately abstract, and
our relatively abstract intuitions are all that we can now properly consider,
without the full mechanism of Multiple-Act Consequentialism and the
Proposed Code. Such abstract intuitions are also least subject to disruption
by highly hypothetical cases. Detailed cases and very speci¬c normative
details, such as that on one option there will certainly be no sentience at all,
can be properly engaged only through the mechanism of Multiple-Act
Consequentialism and the Proposed Code.
Those are also the locations where the value of human life is most
obviously and directly re¬‚ected in the overall account developed here. It
is perhaps a little disconcerting to call murder or death a detail. But the
general signi¬cance of murder and of death is underwritten at a different
place in my overall account than we now engage, through the mechanism
of Multiple-Act Consequentialism. Perhaps, to reinforce this point, it is
worth remembering that there are literatures charging plausibly that aver-
age and total utilitarianism, not to mention negative utilitarianism, also
have murderous or genocidal implications when deployed in conjunction
with act consequentialism,42 but that is not taken as conclusive refutation
of utilitarianism. In light of MAC, it should not be.
The kinds of interaction of the structural elements of HMP that we
can directly confront with commonsense intuitions at the level of gener-
ality and abstraction that is now appropriate point in fact to a con¬‚uence
of HMP and those intuitions. This is partly because of the ¬rst aspect
of HMP: its hedonism. As we have seen, HMP directs a maximin dis-
tribution not over lives but over ¬‚ecks of lives. And it gives pain central
importance. According to HMP, signi¬cant painful injury in an otherwise
very successful life will be more morally signi¬cant than, for instance, a
general deadening of pleasant possibilities in a relatively unsuccessful life.
It is downside risks of that particular sort, risks of signi¬cant pains, that
have special salience given the overall form of HMP. And note that we do

42 Henson (1971); Carson (1983).

intuitively cede to signi¬cant salient risks of such painful physical injuries
and tortures a special normative weight. Torture and other great physical
cruelties and dangers, physical disease and injury and other things closely
related to physical pains, do engage intuitive moral concern in an imme-
diately powerful way. And even salient risks of such things are taken very
seriously in our commonsense evaluations. As we tend to the wounded,
we forget about how rich or poor they are, and focus on their immedi-
ate suffering. And if someone is in immediate danger of being slashed or
tortured, if that is a salient risk, then that is also of great intuitive con-
cern, whatever their overall prospects in life. I have already argued that
a focus on cases of torture is not intuitively misleading in our current
argumentative context.
Of course, there are other sorts of goods, such as life and health and
property, that have intuitive normative weight, and that engage different
intuitions than those that best support my hedonist principle. And some
of the goods I deployed in the last paragraph, like injury per se as opposed
to the pain it causes, are only indirectly though still closely related to the
basic value that HMP re¬‚ects. But we should expect intuitions about those
other goods, given the structure of my proposal, to be fully engaged only
at a different point, through the mechanism of Multiple-Act Consequen-
tialism. We will look at some of the relevant details in the next chapter.
And it is also relevant that we saw in the last part that the hedonic theory
of basic value on which we rely here is itself in fact suitably intuitive.
Let me summarize the main elements of this second reply: The princi-
pal abstract structural interaction within HMP suggests a special normative
salience for risks of certain urgent needs, because pain is closely associ-
ated with intuitively urgent needs. And that is an intuitive feature of my
view. When there is sudden physical and painful danger to someone, it is
intuitively absurd to pause and worry about how their lives are going as
a whole. The intuitively pressing nature of pain, which we also consid-
ered in the last part, undergirds the most salient interaction of the three
structural elements of HMP. While my attempt here to link the quality
of ¬‚ecks with the intuitive urgency of need must be quali¬ed, since some
cases of intuitively urgent need involve death or physical debilities that are
not painful, still those other cases are appropriately and naturally re¬‚ected
in this account by the extended conception of the good incorporated in
the Proposed Code, supported partly by the mechanism of Multiple-Act
Consequentialism. In my second reply, I also invoked the unreality of the
details of some of the obvious intuitive cases that seem to tell most against

We are moving toward a consideration of interactions that also involve
MAC and not merely the various aspects of HMP, and they are the proper
business of the next chapter. But still, there may seem to be other telling
cases of the general sort under consideration now that are not at all unreal
and that cannot be evaded by the mechanism of MAC, and hence are
untouched by my second reply. So let me develop a third reply, in which
considerations deployed in my ¬rst two replies interact, and in which
some new ones come into play.
Consider choices between long-term policies that affect population.
We really have such choices, and they seem unregulated by common-
sense morality and hence by the group agents on which MAC relies to
block unintuitive suggestions of consequentialism. So they may seem to
be plausibly relevant intuitive counterexamples to my proposal.
If applied directly to some population cases, HMP implies instances
of what Par¬t has called “The Repugnant Conclusion”.43 Say there is a
choice between a population of ten billion people, each with a very happy
life, and a larger population involving lives that are barely worth living in
the sense suggested by HMP. HMP implies that the second situation is
better. Indeed, if the choice is between situations without physical pain,
in which there are no ¬‚ecks of negative hedonic tone, then it will always
be better according to HMP to have more individuals with longer lives
and hence more ¬‚ecks. That may seem quite unintuitive.
But, I reply, there are two mitigating factors.44 First, remember that
our commonsense intuition is misled when we consider a case for which
we do not feel full responsibility. And if the choice were actually given us
to eliminate someone in order to increase the quality of other lives, then
we wouldn™t have such a de¬nitive contrary intuition. And the second
mitigating factor is that this particular population case is also unreal in
one crucial way, in fact in a way that suggests some intuitive problems
for HMP from the opposite direction. That is because in reality any
additional persons in the world will inevitably suffer some physical pain.
So other things equal, it seems, any increase in the population would be
bad according to HMP. Remember that it would be better that there be
no one, according to HMP, than that there be anyone who feels or even
risks a ¬‚eck of pain.
This in itself may seem quite intuitively problematic, of course. But
remember that there are also no realistic cases in which we are faced

43 Par¬t (1984: 388).
44 For a classical utilitarian reply to Par¬t, see T¨ nnsj¨ (1998).
a o

with the certain extermination of all sentient life. And in any case, the
extermination of the sentient invokes the mechanism of MAC and hence
the details of the Proposed Code.
Still, if some population policy, or even indeed a group act that con-
stitutes part of commonsense morality, enables the human population to
be larger than it otherwise would be, then the population implications of
HMP may suggest that such an intuitive policy or group act is morally
Because this particular worry applies not just to population policies
but to the crucial group acts deployed by MAC, I will directly discuss it
in the ¬rst section of the next chapter, where we trace the interaction of
HMP, MAC, and commonsense morality. For now, let me simply discuss
a key case that puts some of the apparent population implications of HMP
in a proper and better light.
The real population-based worry about HMP is that it suggests that a
population of zero would be better than any realistic larger population. But
there are other things that many of us believe that suggest that also. Many
of us believe that the world as it is includes suf¬cient evil to show that
a bene¬cent, all-knowing, and all-powerful god does not exist. Indeed,
whatever we moral agents do, our world will likely always be suf¬cient to
reveal that, even if we ignore the ugly history of the world. It is clearly
not the best world there might be, according to HMP or any plausible
principle for ranking worlds. But the theist can plausibly remind us that
there might not be a best possible world. It might always be possible
to create another world that is better than any given world. And if we
reply that a world created by a benevolent being would yet need to be
better than this one, there is yet a reasonable question about how good
a world would really have to be to reveal the bene¬cence of a creator, if
there is no best of all possible worlds. So what, then, is our complaint?
Surely we have one. The real problem of evil that many of us feel, it
seems, is that the world in fact is worse than nothing. Remember that the
true evaluation of outcomes is revealed in conditions of full responsibility
for those outcomes, and it plausibly seems to Ivan Karamazov that any
god who would erect a fabric of human happiness on the torture of one
child would be morally problematic. And so the creation of our world, in
which many children are tortured each day, would have been, many of us
think, morally problematic. Remember also that traditional theological
responses to the problem of evil often invoke an implausible afterlife that
is supposed to make up for the earthly suffering of the innocent. And
recall that in our discussion in the ¬rst part of this chapter we saw that

suitably re¬‚ective intuition suggests that even a happy afterlife could not
fully recompense the intense suffering of a child.
Our real world is in fact worse than nothing, even according to much
less pessimistic principles than HMP. So our contrary intuitive reaction
to the key population implication of MAC does not really undercut the
pessimism of HMP. That intuition is distorted. It fails to acknowledge
pretty obvious facts about the overall value of the world, facts that don™t
even drive deep enough to engage the particular pessimism of HMP.
What may seem the central counterintuitive implication of HMP is in fact
true, though it is hard not to avert our happy eyes from that unpleasant
Still, some may insist, while much of the world is dark, it is not immoral
to have a child and hence to create another life, or to extend your own,
even though the child™s life and the extension of your own will surely
involve some pain. But HMP does not imply that all realistic individual
lives are on balance bad for the whole. Any life or segment of a life will be
on balance a good thing if the most intense pain it involves is less than the
most intense pain of others that it assuages. It is easy to have a worthwhile
life, though that does require that you do something for suffering others.
We must consider yet the detailed interactions of HMP and MAC.
But insofar as we can confront HMP and abstract and properly recti¬ed
commonsense intuition directly, we may conclude this chapter in this
way: We have seen that a distributional focus on moments or ¬‚ecks or
something close is appropriately intuitive. We have seen that maximin is
appropriately intuitive, since after Part One we can properly assume that
there isn™t a deep moral difference between harming and not aiding. We
have seen that risk aversion is not intuitively inappropriate, at least if the
Proposed Code is suitably intuitive in detail. And we have seen that the
abstract structural interaction of these three elements is at least roughly
intuitive in the further context set by what we already know to be the
hedonism of our principle and the properly intuitive form of Multiple-Act
Consequentialism, at least if the Proposed Code is properly intuitive, and
if the one-off group agents that it deploys are in fact suitably bene¬cent
according to HMP.
Still, we do face another abstract intuitive worry about the interactions
of the various elements of my proposal. That involves a key interaction
of MAC and HMP. In Part One, I defended the intuitive plausibility of

45 1 have company in these sorts of claims. See Fehige (1998: 521“523).

MAC by appeal to the plausibly bene¬cent consequences of various group
acts. But I did that before I developed the somewhat unconventional value
theory expressed by HMP. I have ordered the book in this way so that
I move sequentially from the least to the most provocative elements of
my proposal. But it is reasonable to worry that my earlier arguments in
defense and support of MAC will look considerably more problematic in
light of HMP. So we will need to reexamine those arguments in the next
chapter. I have already promised to examine the population implications
of the relevant group acts anyway.
Nevertheless, we can now conclude that we do not know on the basis
of relatively abstract commonsense intuition that HMP alone is false in its
structural features. And we can plausibly conclude that its three structural
elements are indeed positively supported by properly re¬‚ective intuition,
if the Proposed Code implied by a conjunction of HMP and Multiple-
Act Consequentialism in our world has an intuitive detailed form. We are
almost home.

Part Four
Advice for Atomic Agents
A Code

The conjunction of Multiple-Act Consequentialism and the Hedonic
Maximin Principle modulates a traditional utilitarian concern about
hedonic good through a focus on two sorts of justice: distributive jus-
tice and the avoidance of unjust means such as murder. I will refer to this
marriage as the just good theory, or JGT.
JGT has been directly vindicated on the basis of ordinal hedonic value,
and we have seen that each of its elements is intuitively appropriate at an
abstract level, subject to some reservations about how the details work.
We face one more argumentative burden. This chapter argues that, in
the world we inhabit, the just good theory supports a moral code that I
will call the Proposed Code, or PC for short, and that PC is a properly
intuitive detailed morality.
PC is a relatively concrete set of normative directions. I will argue that
it is within the range of contemporary re¬‚ective normative intuition of
the same level of detail. You easily may have forgotten how JGT, which
is a form of direct act-based consequentialism, can support a ¬xed moral
code at all. The answer is that it supports such a code in the context of
the actual group acts in which we participate. It is a code for us in those
somewhat contingent circumstances.
PC has two major components, which re¬‚ect major components of
the just good theory and also of contemporary reconstructions of com-
monsense morality. First, there is a conception of the good or well-being,
the focus of section II. Second, there is a conception of duty and virtue,
the focus of sections III and IV. Section I articulates various presump-
tions of our discussion, and section V considers the differences that total
utilitarianism would introduce.


We must trace here the detailed con¬‚uence of commonsense morality
and the just good theory. Commonsense morality makes a number of
presumptions that can appropriately narrow our focus.
First, commonsense morality presumes that there is a single morality
that is appropriate to the condition of at least the range of readers of
a book like this. Application of MAC naturally takes the form of rec-
ommendations about reasons for atomic agents to accept. But our ¬rst
commonsense presumption implies that this advice not only must be sin-
gle and coherent for a particular atomic agent, regardless of the many
overlapping agents in which it takes part at a moment, but also must be
generally applicable across people in our roughly contemporary and local
conditions. So we will seek a single and relatively general code of properly
accepted reasons for atomic agents like us. That is the kind of code with
which we can confront common sense.
Second, commonsense morality presumes that the background con-
ditions of our lives “ for instance, the existence of friendships, families,
and property, and the relatively contingent features of our psychologies
required if we are to accept reasons “ are at least roughly appropriate.
They might be appropriate if they are natural, inevitable, and unalterable,
and hence not really optional. Or they might be appropriate if, though
optional, they are legitimate in some adequate normative sense. I don™t
con¬dently believe either of those things. I have already registered wor-
ries about the particular case of the practices of risk sharing that underlie
our intuitive reactions to the Trolley Problem. But any reasonably close
con¬‚uence of any ethical theory and commonsense morality can plausi-
bly be sought only within the bounds of this presumption. It may seem
that such background conditions of our lives would be more problem-
atic according to the highly risk-averse and distribution-sensitive HMP
than according to more familiar normative principles for ranking out-
comes. But while that issue is too complex to treat fully here, we will
return in a moment to analogous and more pressing possible differences
between HMP and familiar normative principles, during discussion of
my eighth claim, and see that they are less dramatic than they might ¬rst
Third, commonsense morality presumes that any moral code will have
a certain abstract form. PC will be not only an articulation of appropriate
reasons, but also an articulation that constitutes distinct conceptions of
the good, of virtues, and of duties and obligations. Some aspects of that

traditional organization are theoretically underwritten by JGT, but the
necessity of the full traditional form is not obvious. Still, such a form
seems largely a matter of emphasis and organization, and will aid in our
attempt to gauge the con¬‚uence of the just good theory and commonsense
morality. Otherwise, things are too complex to handle tractably.
Our fourth presumption is a commonsense conception of our options
and alternatives and of the attendant salient risks. We can plausibly seek
a con¬‚uence with commonsense morality only within this constraint.
Perhaps by suf¬cient cleverness we could make the world an Eden for
everyone, and in a way that would be very morally salient according to
JGT. But I will put such radical possibilities aside here.
A ¬fth presumption is less within than about common sense. Our
actual morality is quite complicated, and the various normative theo-
ries rooted in commonsense intuitions, and also various philosophers™
attempted reconstructions of commonsense morality, are perhaps over-
simpli¬ed. But I will presume that philosophical reconstructions and intu-
itionisms are at least roughly correct characterizations of commonsense
morality. We will cleave close to the detailed range of intuition-based nor-
mative theories and attempts to sketch commonsense morality that have
been developed by philosophers, as if they were each at least roughly ade-
quate reconstructions of local commonsense morality. They are, I believe.
This presumption implies, along with the variety of reconstructions and
intuitionisms re¬‚ecting even contemporary common sense, something
that might also be expected independently. Our commonsense morality,
even if we focus relatively locally on our own morality now, is indetermi-
nate, vague, and incoherent in various ways. That makes it easier for JGT,
since it need underwrite only a speci¬c coherent code somewhere within
that range. But while some philosophers are con¬dent that their own
detailed and idiosyncratic normative intuitions capture common sense,
while many of their close colleagues™ do not, that is very unlikely. Still,
I take the test of common intuition to be more restrictive than some
do. I think that the just good theory is false if it has to bite too many
The sixth constraint on our project here, of course, is JGT. It is the
conjunction of the Hedonic Maximin Principle and Multiple-Act Con-
sequentialism. Let me recapitulate. The Hedonic Maximin Principle is
this: First, of two lotteries over feasible worlds that consist of the same
number of equally probable outcomes, the better lottery is the one that
has the better worst-possible outcome. If they have equally bad worst-
possible outcomes, then the better lottery is the one that has the better

second-worst-possible outcome, and so on. Given the method developed
in Chapter 6 for turning lotteries with different numbers of differently
probable outcomes into equiprobable lotteries with the same numbers
of possible outcomes, this yields a complete ordering of lotteries over
worlds, which implies that of any two lotteries over worlds, the better
lottery is the one that has the better worst-possible outcome. Second, of
two feasible worlds that contain the same numbers of ¬‚ecks of experi-
ence, the better world is the one that has the better worst ¬‚eck. If they
have equally bad worst ¬‚ecks, then the better world is the one that has
the better second-worst ¬‚eck, and so on. And any world is equivalent
in value to another that has the same number of ¬‚ecks at each level of
positive or negative ordinal value as the ¬rst plus any number of ¬‚ecks
of null value. So this provides a complete ordering of feasible worlds.
According to Multiple-Act Consequentialism, atomic agents accept
reasons that constitute, govern, and balance an overlapping multiplicity
of group agents. MAC applies a basic normative principle like HMP to
rank available options. It holds that the most choiceworthy is the best.
But this direct evaluation is performed all at once on the options of all
the multiple agents that overlap in a given atomic agent. There will be
practical con¬‚icts among the various agents that overlap. In a case in
which an atomic agent can defect from a group agent with a proper
project, and hence achieve additional good consequences on the side
while not undercutting the project of the group agent, we are to compare
a ¬rst counterfactual situation in which the atomic agent achieves what
it can by the defection but in which the various other atomic agents that
constitute the group agent do not constitute such an agent, to a second
counterfactual situation in which the group agent acts as it does and the
atomic agent does not defect. If the ¬rst situation is better, then MAC
says to defect. If the second situation is better, MAC says not to defect.
That is the principle of Very Little Defection, or VLD. In a case in which
we must assess the relative importance of two forms of overlapping but
con¬‚icting group agency for an atomic agent that is part of both, we also
compare two counterfactual situations: In the ¬rst situation, the ¬rst group
agent doesn™t exist because the atomic agents in question fail to properly
constitute such an agent, but the second group agent has its actual form.
In the second situation, the second group agent doesn™t exist, but the ¬rst
has its actual form. If the ¬rst situation is better, then the second group
agent is more normatively signi¬cant, and MAC says to defect from the
¬rst group agent in favor of the second. This is the principle of Defect to
the Dominant, or DD.

Our seventh and crucial presumption is that there are in fact the
group agents that I will deploy in my remaining remarks in this chapter.
Remember my argument in Chapter 2 that our cooperative norma-
tive practices are characteristically one-off group agents, supported by
entwined reciprocity and bene¬cence.
I will also repeatedly rely here on an eighth claim, that these cooperative
normative practices are proper by the strictures of HMP. That may seem
wrong, since HMP is unusually distribution-sensitive and risk-averse. And
remember that I defended the intuitive cogency of MAC in Chapter 3
by appeal to the intuitively bene¬cent consequences of various group
acts that constitute our deontic moral practices. So if HMP ranks these
acts in a different way than more familiar normative principles, those
arguments may need to be revisited. On this matter, I shouldn™t rest on
mere presumption. So let me make three replies to various aspects of
this objection.
One part of the worry can be put this way: VLD, incorporated in MAC,
implies that we shouldn™t defect from a group act with good consequences
unless by defecting we can achieve more good than the entire group act
achieves. And I claim that many of our deontic practices are group acts
with good consequences, from which we should rarely defect. But while
this may seem suitably plausible if we presume a familiar utilitarian basic
normative principle, HMP seems quite different. It suggests that we focus
on the worst risks of intense pains that attend the group act and the
individual act of defection. And since group acts have a considerably
greater impact on the world than individual acts, they may hence seem
to carry a much greater risk of harm. MAC may seem always to counsel
defection from group acts when conjoined with HMP.
But, I reply, in a situation where defection is at issue, according to
MAC we need to consider not merely the risk attendant on the group
act in question, but the risk attendant on its absence. And because it is a
group act, both risks are correspondingly large. According to VLD, we
compare, on one hand, the world with the defecting act and without the
group act, and, on the other, the world without the defecting act but with
the group act. Even if the positive act of defection in itself risks little, the
absence of the group act might risk a lot.
Still, it may seem that our moral practices generally, and hence the
group acts on which I rely here, may be problematic according to HMP.
They may increase some risks of great local evils. And recall our truncated
discussion of population issues in the last chapter. If our common moral
practices aid the stable survival of very large societies, if they allow the

human population to be much larger than it would be in some alternative
state of nature, then the pessimism of HMP may suggest that they are
inappropriate group acts.
But, I reply, in the cases under consideration the pessimism of HMP
is focused through the mechanism of MAC on the worst risks for intense
pain presuming, on one hand, individual defection from a group act and
the absence of the group act versus, on the other hand, the worst risks for
intense pain of the group act itself. And that helps with part of this worry.
It is undeniable that some very large cooperative group acts “ for instance,
robust mass movements “ may be quite relevantly dangerous according
to JGT. They may increase the risk of organized and effective torture,
for instance. It might risk less intense pain if they did not exist. But the
deontic moral practices on which I rely here are one-off group acts that
characteristically alleviate individual suffering. The risk of their misuse is
not that of some massive and generally capable group agent that can be
de¬‚ected toward evil ends.
There are possible worries about this reply. Apparently Goebbels com-
plained that the Italians lacked suf¬cient honesty to effectively administer
the Final Solution. Perhaps our one-off practice of not lying is a mecha-
nism that can be put to a variety of evil uses. But, I reply, the common
moral practice of refraining from lies involves plausible exceptions for
disasters. A more general worry is that if there are unforeseen interac-
tions between even a group act that is inherently bene¬cent and various
contingencies, the group act can end up not only risking but also achiev-
ing an unforeseen bad end. But the group acts constituting our standing
moral practices are so vast that almost all relevant eventualities and remote
likelihoods will in fact show up across their range, and would show up
across the vast range of their absence. There are no relevant unforeseeable
To summarize the ¬rst two parts of my reply: There are great risks
that attend any group act, but so too its absence. And the general and
inherent nature of the very large one-off group acts which constitute our
common moral practices is to weaken risks of intense pain. If life threatens
to be nasty, brutish, and short without commonsense morality, then HMP
would be very responsive to that increased nastiness.
There remains the worry about populations, the worry that because the
group acts that constitute commonsense morality support larger human
populations, they are contrary to HMP. But the sensitivity of HMP to
locally intense nastiness would trump its limited pessimism about larger
populations. One torture is more signi¬cant than lots of mild pain. And in

any case there is the next point, my third reply. Recall that the group acts
relevant to MAC are timeless. They include actual future cooperators. So
to a large degree relevant populations are presumed ¬xed in the evaluations
relevant to MAC. All of the population inside a group agent constituting
a moral practice will be held ¬xed in the key evaluations by MAC of
defection from that agent.
We face worries not only about the bene¬cence of the group acts that
constitute commonsense morality, but also about the population polices
that HMP suggests. At the end of the last chapter, we noted some factors
that mitigate these worries. But the three replies we have just discussed
also suggest that when population policies constitute widespread one-off
group acts, they should be more optimistic than HMP alone may suggest.
With all presumptions and clari¬cations now in place, we face two
questions: Given a commonsense view of our salient alternatives and
risks, assuming the legitimacy of the background institutions in which
our lives take place and that help to form our psychologies, speci¬cally
including the bene¬cence of the group acts that constitute many of our
moral practices, and presuming a familiar code form that involves a con-
ception of the good, of virtues, and of duties and obligations, what reasons
should be accepted by all atomic agents in our more or less contempo-
rary circumstances according to the just good theory? PC is the answer
to this question. Our second question is whether PC is within the range
of contemporary re¬‚ective commonsense morality that is revealed by the
range of plausible intuitionisms and other philosophical renditions of that
commonsense morality. To pre¬gure, it is.


The ¬rst element of PC is a speci¬cation of the good or well-being.
One conception of the good natural to PC is of course the conception
of fundamental good sketched in section I of Chapter 3. First, there is
basic hedonic good, enshrined in HMP. Second, Multiple-Act Conse-
quentialism implies that other sorts of good are also fundamental. In the
¬rst place, there are agency goods, conditions necessary to the existence
of effective agency, which when in the service of proper projects become
genuine agency goods, and hence one sort of fundamental, though not
basic, good. These include true belief, life and health, physical and men-
tal abilities, freedom from injury, and freedom from imprisonment and
manipulation, when those things are in the service of proper projects. And
there are also other ways in which agency goods can become fundamental

goods that are neither “basic” nor “genuine”, when they are protected
by one-off group agents that are properly bene¬cent and yet that extend
respect to agency goods that are not in the service of proper projects.
There are other sorts of goods that, while not quite fundamental, yet
are entwined in characteristic close ways with fundamental agency goods
or basic goods. This suggests a relatively central role even for these goods
within PC. They have stable but contingent instrumental relationships
with fundamental goods. I will call these tertiary goods.
There are tertiary goods closely related to agency goods. Social and
economic power and position are goods that extend intuitive capacities
for agency, and as a matter of general but contingent fact are relevant
to our relatively general code. These are what I will call genuine tertiary
goods only when they support fundamental agency goods. In other words,
two tests must be met by genuine tertiary goods. They must have stable
instrumental relationships of the right sort, as a matter of general fact. But
they also must have in their particular case a proper relationship to the
right sort of project in fact.
There are also tertiary goods that are closely related to basic good.
Frustration of desire is sometimes painful even in the very concrete sense
invoked by HMP. Remember our quasi-experience of physical pain.
Certain general forms of desire frustration probably have enough of even
a stable relationship to pain to count in some instances as genuine tertiary
goods, to meet our twinned tests for such goods.
There are plausibly also cases in which tertiary goods “ for instance, the
absence of frustration or having wealth “ are important but not genuine.
This status can be delivered by the same indirect mechanism that generates
analogous fundamental agency goods.
There are also more complex entwinings due to instrumental relation-
ships that may be relevant. For instance, agency goods are general means
to the satisfaction of the agent™s desires. And sometimes stable instrumen-
tal relationships reinforce other sorts of status as a good. For instance,
many forms of death or illness are also painful for the victim in a suitably
general way.
In summary, PC includes a three-level conception of good or well-
being, which I will hereafter call TLC (for three-level conception), and
which incorporates (i) basic good “ positive hedonic value and the absence
of negative hedonic value; (ii) fundamental goods, including both gen-
uine agency goods “ true practical belief, life and health, physical and
mental abilities, freedom from injury, and freedom from imprisonment
and manipulation “ when these things support proper projects, and also

those agency goods not in service of proper projects that are yet protected
by bene¬cent one-off group acts; and (iii) important tertiary goods “
which include genuine tertiary goods like wealth and position that sup-
port proper projects and desire satisfaction linked with hedonic value, and
other such goods that are yet protected by proper projects. Agency goods
that are not fundamental and tertiary goods that are not important might
be said to be goods in an extended though not normatively relevant sense.
Within even this three-level conception, the class of basic goods is pri-
mary in an important sense, and tertiary goods in third place, because of
the structure of JGT.
That is a sketch of the conception of the good to be included in PC.
Of course, there are complications. I™ve already said that agency goods
are genuine only when in service of proper projects; likewise for tertiary
goods. But notice that this implies that the conception of the good rests
on a conception of proper projects. As noted in Chapter 2, there are
complexities regarding the general nature of proper projects beyond the
obvious fact that they are at least roughly in accord with HMP. And TLC
suggests an obvious extension of our initial notion, so that proper projects
in a secondary but real sense also include those increasing fundamental
agency goods and important tertiary goods.
TLC would seem to require some way to balance off various possible
increases and decreases in agency goods or tertiary goods. But notice that
even the primary normative balancing delivered by HMP has not yet
been incorporated into the conception of the good I have sketched here,
at least in any direct way. We have not yet developed a speci¬c conception
of when one thing of one type is better than another of another type;
we have only a vague conception of the elements of the good in a rough
normative priority suggested by JGT.
There is another large and related complexity. Basic hedonic value is
the primary good of the primary moral patients, the primary objects of
moral concern, which includes all the suitably sentient. But we saw in
Part One that MAC suggests that some humans, those with whom we
cooperate in bene¬cent projects or who engage in bene¬cent projects on
their own, have a second status as another sort of moral patient. They
are owed different types of respect. What™s more, other forms of respect
for agents beyond the group may be mandated by bene¬cent group acts.
And notice that the existence of group agents suggests that some agency
goods accede to groups themselves. Some organizations are richer or more
powerful than others. When pursuing proper projects, even group agents
become moral patients in a secondary sense. This status, as a matter of

contingent but relatively stable fact, accrues only to things composed of
atomic agents that are moral patients in the primary sense. But obviously,
distributions of goods to one sort of moral patient can con¬‚ict with those
to another.
So there are various complications that involve trade-offs and balances
among goods that PC has not yet addressed. The structure of JGT naturally
suggests that such con¬‚icts and balances should ultimately be assessed by
HMP, and our discussions of MAC have given us some additional mech-
anism to that end. But for the moment I will ignore these complexities,
and leave TLC vague. That is for two reasons. First, it is largely through
its deployment mediated by conceptions of duty and virtue, to which we
will shortly turn, that this conception of the good is practically signi¬cant.
In other words, proper consideration of balances and trade-offs will be
our concern in sections III and IV. Second, the commonsense concep-
tion of the good is hardly more determinate than my vague conception
in these ways. We can see this if we turn to a confrontation of TLC with
common sense.
My ¬rst argument that this ¬rst element of PC is suitably intuitive is
simply that it closely tracks the range of reasonably intuitive views of the
good that have been developed by and are still discussed by philosophers,
and that we considered in Chapter 4. Hedonism is one classical view, but
so are desire-satisfaction views, and also objectivist views that make much
of what we know now to be agency goods. And that suggests the very
range of goods recognized by PC.
There are complexities. First, wealth and power are not parts of most
philosophers™ conceptions of the good. But they are, on the other hand,
very intuitive goods. Second, only agency goods entwined with proper
projects are genuine goods according to PC, and this may seem at least
somewhat unexpected. But it does not violate any clear and certain com-
monsense intuition. For instance, TLC is more moderate in this respect
than Kantian accounts that would grant the happiness of the vicious no
basic value at all. And of course, some agency goods are fundamental but
neither basic nor genuine.
It is also relevant, as we saw in Chapter 4, that the various con¬‚icts
among the differing philosophical accounts of the good are best resolved,
and on intuitive grounds, in the direction of hedonism, a fact that is
re¬‚ected in the basic status of hedonic value even in PC. We saw in Chap-
ter 4 that hedonism alone is adequately intuitive, adequately within the
range of recti¬ed commonsense intuition. But the conception of good in
PC merely organizes in a certain manner the basic elements recognized

by all traditional competitors to hedonism and also by common sense,
a manner that re¬‚ects the properly dominant role of hedonic value. It
captures more of the subtleties and con¬‚icting tendencies of our com-
monsense conception than hedonism alone, though we saw earlier that
hedonism is, even on its own, at least adequately intuitive.
My second argument that TLC is properly intuitive is its close con-
¬‚uence with Scanlon™s recent sensitive and intuition-based account. He
notes that any plausible theory of well-being would have to recognize at
least the following ¬xed points:

First, certain experiential states (such as various forms of satisfaction and enjoy-
ment) contribute to well-being, but well-being is not determined solely by the
quality of experience. Second, well-being depends to a large extent on a person™s
degree of success in achieving his or her main ends in life, provided these are
worth pursuing. This component of well-being re¬‚ects the fact that the life of a
rational creature is something that is to be lived in an active sense “ that is to say,
shaped by his or her choices and reactions “ and that well-being is therefore in
large part a matter of how well this is done “ of how well the ends are selected
and how successfully they are pursued. Third, many goods that contribute to
a person™s well-being depend on the person™s aims but go beyond the success
of achieving those aims. These include such things as friendship, other valuable
personal relations, and the achievement of various forms of excellence, such as in
art or science.1

TLC recognizes all these elements, with a certain characteristic spin. Cor-
responding to Scanlon™s ¬rst ¬xed point, we have basic hedonic value.
The spin is that hedonic value is what is involved in the relevant expe-
riential states. Corresponding to Scanlon™s second ¬xed point, we have
proper projects helping to constitute genuine tertiary and agency goods.
The spin is that what is worth pursuing is proper projects. Correspond-
ing to Scanlon™s third ¬xed point, we have agency goods, and proper
projects relevant to certain forms of group agent. The spin is that friend-
ships and personal relations aiming at evil are not genuine goods, though
they may be fundamental. Certainly this spin on Scanlon™s intuitive ¬xed
points is well within the range of determinacy of our re¬‚ective common
Scanlon is an important recent non-consequentialist foil to whom we
will return in our discussion of duties and virtues. But he is also an
immediately useful foil in two other ways. First, it may help to soften

1 Scanlon (1998: 124“125).

some resistance to the ¬rst element of PC to note that, despite my
consequentialism, I agree with Scanlon™s theoretical contention that well-
being and the good are not “master values”. I agree that the conception of
one™s good or well-being doesn™t play much role in one™s individual prac-
tical reasoning, though its components often do, so that it can be a kind
of translucent Trojan horse: While it doesn™t do much work beyond that
which its obvious contents do on their own, yet it can introduce a cer-
tain element of distortion. And I also agree that for different third-person
distributional judgments in various contexts, different detailed forms of
well-being are normatively salient.2 I don™t suggest that the post of¬ce
consult TLC to determine where to deliver a piece of mail. Still, despite
the fact that the good and well-being are not “master values”, a rough
conception of individual well-being or good like TLC, or indeed Scan-
lon™s relatively similar alternative, does play a general and necessary role
in underwriting and specifying the duties and virtues that we will shortly
Scanlon is also an immediately useful foil in another way. PC speci¬es
a set of accepted reasons. So we face the question of how proper reasons
as speci¬ed by PC can encompass a conception of the good such as the
one we have just traced. Part of the answer is that it is refracted through
the duties and virtues we will soon discuss. But there is another part of
the answer, which we can take from Scanlon.3
Scanlon contrasts his view on this matter with what he calls the tele-
ological view of value, the view that what “we have reason to do . . . (at
least as far as questions of value are concerned) . . . is to act so as to realize
those states of affairs that are best “ that is, have the greatest value.”4 He
complains, against the teleological view, that we value friendships and
worthwhile activities from the inside in a different way, as things to be
respected. This suggests a broader account of value.
To value something is to take oneself to have reasons for holding certain positive
attitudes towards it and for acting in certain ways in regard to it. . . . They generally
include, as a common core, reasons for admiring and respecting it, although
˜respecting™ can involve quite different things in different cases. Often, valuing
something involves seeing reasons to preserve and protect it (as, for example,
when I value a historic building); in other cases it involves reasons to be guided
by the goals and standards that the value involves (as when I value loyalty); in

2 Ibid., 108“143.
3 Ibid., 78“107.
4 Ibid., 80.

some cases both may be involved (as when I value the U.S. Constitution). To
claim that something is valuable (or that it is “of value”) is to claim that others
also have reason to value it, as you do.5

This seems at least roughly correct. To have a conception of the good
such as we have sketched here is at least in part to accept reasons to
admire and respect basic goods, fundamental agency goods, and important
tertiary goods, in the relevant ways that Scanlon notes. As Scanlon would
not deny, this includes in some cases familiar teleological respect. As he
probably would deny, this teleological respect has the kind of organizing
role suggested by the primary status of hedonic value in PC. But, as I said,
we will return to these issues of trade-offs and balances of goods shortly.
There is also a second layer in this translation between conceptions of
the good and accepted reasons. To have such a conception of the good is
also to accept higher-order reasons that specify that one should accept the
¬rst-order reasons, whoever one is. This way to treat claims to universality,
such as that made in the last sentence of the quotation from Scanlon, is
familiar from Allan Gibbard™s work on the acceptance of norms.6 One
¬nal complexity: PC must recommend acceptance of these reasons as
agent-governing reasons, which govern the kinds of projects agents are
to adopt. But it also must recommend acceptance of these reasons at
a lower order, as the kind of agent-constituting reasons that constitute
individuals and groups to have in fact certain sorts of projects rather than


The second component of PC is a speci¬cation of virtues and duties. It
will be tailored for comparison to commonsense morality.
There are a number of axes on which plausible coherent reconstruc-
tions of our diverse and marginally coherent commonsense morality may
reasonably differ. We can develop at least a rough sense of plausible posi-
tions on these axes by an historical survey of detailed formulations, recon-
structions, or reformulations of commonsense morality by philosophers
and religious ¬gures, beginning with views that helped to form our nor-
mative tradition, and then moving on to more recent views that re¬‚ect
it. My discussion re¬‚ects three historical waves of such reconstructions.

5 Ibid., 95.
6 Gibbard (1990b: 153“170).

I will not here attempt a general summary of these accounts, but they
will inform the axes of difference that we will later consider. While my
dominant concern is the contemporary commonsense intuition of West-
ern ethicists, the historical elements of this survey will allow us to consider
some of the various tendencies our common sense inherits, and will also
serve to suggest a broader possible consensus.
The ¬rst historical wave consists of four religious documents. First,
there are the nonreligious commandments of the Decalogue.7 Second,
Buddha™s famous sermon in the Deer Park at Benares includes a cru-
cial statement of his Middle Path, rooted in the Four Noble Truths that
include a speci¬cation of the Eightfold Way to the cessation of craving.8 In
China, there is an important Buddhist tradition, and the Taoist tradition
also incorporated a revisionary normative element allied with the Bud-
dhist extinction of craving.9 But Confucian doctrine developed a third
tradition of moral theory, which is easier than Taoism to bring into contact
with our main concerns. Whereas the Decalogue is deontological, and
Buddhist ethics perhaps largely deontological though with some virtue-
like elements, the Analects of Confucius suggest a largely virtue-based
account, leavened with some deontological elements.10 Our fourth reli-
gious document, the Qur™an, includes nonreligious ethical formulations
that seem evenly deontological and virtue-based.11
The classical tradition represents a ¬fth alternative to the four reli-
gious accounts. But it is best represented by philosophers. Our second
historical wave encompasses the three detailed reconstructions of com-
monsense morality by world-famous Western philosophers, by world-
historical philosophers of great in¬‚uence whose works are taught in
standard normative theory courses. These are Aristotle™s account in the

7 Exodus 20: 12“17, King James Version. These were traditionally thought “ for instance, by
Maimonides and Aquinas “ to be subsumed under a single general principle, though there
are some differences about which general principle. See, for instance, Donagan (1977: 57)
and Luke 10: 25“37. It is interesting that Matthew 5: 21“22 suggests apparent extensions
that seem quite revisionary. But even the tenth commandment seems to be a revisionary
8 Buddha (1955). See also Pratt (1928: 59“64). The fourth step of the Eightfold Way is right
action. This is generally sel¬‚ess and charitable, but more speci¬cally circumscribed by the
Five Precepts, which bind lay people as well as monks. See Smith (1958: 119).
9 Lao-Tzu (1993).
10 Confucius (1979). Particulary relevant are Analects IV, 14; VI, 30; XII, 1; XII, 22; XIV, 4;
and XV, 24. For background, see the introduction by Lau in Confucius (1979) or Smith
(1958: 179“191).
11 Arberry (1955) or Dawood (1974). Particularly relevant are Sur¯ h 70 (The Ladders), 22“35;
and Sur¯ h 13 (Thunder), 20“23.

Nicomachean Ethics,12 Aquinas™s in Summa Theologica,13 and Kant™s in the
relatively neglected but also relatively detailed The Metaphysics of Morals.14
This second wave to a degree re¬‚ects its contemporary common sense,
but also helped to form ours.
The third historical wave consists of the detailed reconstructions of
commonsense morality that have been proposed by more or less contem-
porary philosophers, which is to say, by philosophers who may re¬‚ect
our roughly contemporary common sense, but who haven™t formed it.
It encompasses the reconstructions of Bernard Gert,15 Alan Donagan,16
W. D. Ross,17 and Henry Sidgwick,18 and the suggestive partial recon-
structions by Charles Fried19 and Thomas Scanlon.20
Each of the third-wave reconstructions, and to a large degree each of
the second-wave reconstructions, captures in one coherent way various
normative tendencies in our complex contemporary common sense, a
common sense that is the complex heir to the tradition of the Deca-
logue and each of the second wave reconstructions, and to much more
besides. None is obviously false to common sense. It is somewhat fashion-
able to deploy piecemeal deontological intuitions as decisive negative tests
against moral theories, and they may be deployed against all of these recon-
structions. But that would be inappropriate. That is in part because our
commonsense intuitions are in obvious tension in various ways, between
individuals and within individuals. And that in turn is in part because
there are con¬‚icting normative traditions that our common sense inher-
its, and also real normative tensions between intuitive duties and virtues
and between goods achieved by individual and group action. These ten-
sions must be resolved in some coherent way by a coherent ethical theory.
There is no assurance that individual intuitions deployed piecemeal against
a particular theory will not of necessity be abandoned when forced into a
coherent and re¬‚ective system with other intuitions. So the only test that
even our re¬‚ective normative intuition can provide spans the range of the
third-wave coherent reconstructions, and probably also the second-wave

12 Aristotle (1980).
13 Aquinas (1915“38).
14 Kant (1996b).
15 Gert (1998).
16 Donagan (1977).
17 Ross (1930: Chapter II, 16“47).
18 Sidgwick (1907: Book III, 199“361; and Book IV, Chapter III, 423“459).
19 Fried (1978).
20 Scanlon (1998).

reconstructions. I will argue that PC is in the rough middle of the space
encompassed by the third-wave and second-wave reconstructions.
We will now survey various axes of difference among the plausible
reconstructions of commonsense morality, and locate PC along each axis,
concentrating ¬rst on matters of general structure, and then on matters of
speci¬c content. We will take all the threads that comprise JGT, and run
them together through various axes on which the various reconstructions
differ. Or, to put it the other way around, we will develop a new recon-
struction of commonsense morality recognizably related to the others, but
with an eye to the theoretical framework of JGT. We will hence end up
with a detailed code that is well within the range of our contemporary
re¬‚ective common sense. That is the Proposed Code, or PC.
The ¬rst set of axes trace differences among the ways in which the
reconstructions deploy very broad normative categories. Some recon-
structions are primarily duty-based, some primarily virtue-based, and
some mix those resources. Some recognize duties to oneself, and oth-
ers do not. Some recognize special as well as general obligations; others
do not.
On these matters, the general framework of MAC developed in Part
One, at least in the context of our commonsense presumption about the
proper general form of any moral code, implies that PC will occupy a
middle-ground position. It will recognize all of these things “ virtues
and duties, duties to self, and special and general obligations. First, at
least some intuitive virtues and duties to self support one™s own proper
agency over time, because continuing action in pursuit of proper projects
is a kind of group activity in which various temporal segments of one™s
life cooperate. Some agent-constituting, agent-governing, and indeed
agent-balancing reasons enjoin respect for such virtues, in the sense of
“respect” we adopted from Scanlon in section II, and such self-regarding
duties are themselves agent-constituting reasons. Second, special and gen-
eral obligations are also intuitively captured, at least in general form,
by MAC. Group agents with proper projects normatively constrain the
atomic agents that make them up. Proper agent-constituting reasons help
to create one-off group acts that are forms of normative cooperation
supporting some general duties. Such reasons also help to create group
agents with other proper projects that engage or support special obliga-
tions. And proper agent-governing reasons also support those types of
special and general obligations. All this is familiar from Part One. In addi-
tion, proper respect for genuine agency goods and tertiary goods already
incorporated in PC requires that there are agent-governing reasons that

support some general constraints on behavior toward others with proper
A second cluster of structural axes re¬‚ects the various mechanisms that
reconstructions deploy to treat con¬‚icts among duties or virtues. There are
commonsense tensions between commonsense duties. Coherent accounts
must resolve these in coherent ways, and different reconstructions use
different mechanisms to do this. Kant distinguishes perfect and imperfect
duties, where perfect duties can never be violated but imperfect duties pre-
scribe goals too broad and imperfectly attainable for their full achievement
to be absolutely required, and where negative duties, duties to refrain, are
characteristically more perfect. Gert stresses negative duties, with even
positive duties to act being duties to prevent harms, and also allows for
exceptions. Fried suggests that there are absolute or categorical prohibi-
tions against lying and doing physical harm to an innocent person, which
re¬‚ect more general individual negative rights to life, liberty, and property.
He thinks there is also a limited positive duty for concern and bene¬cence.
This is made consistent by his adoption of a very stringent Doctrine of
Double Effect. The DDE suggests that we are specially responsible for
the intended consequences, but not for the foreseen but unintended con-
sequences, of our actions. Another analogous intuitionist mechanism is
Kamm™s Principle of Permissible Harm, that it is permissible for greater
good to produce lesser evil.21 Donagan™s reconstruction rests on the
Pauline Principle, Paul™s suggestion that evil should not be done that good
may come of it.22 He also deploys tacit exceptions, a limitation of positive
duties to otherwise permissible acts, and a complex mixture of nega-
tive and positive duties. Ross grants all duties, negative and positive, mere
prima facie status, but with the prima facie dominance of non-male¬cence
over bene¬cence. Sidgwick claims that ordinary duties are vague and
con¬‚icting, and that the vagueness and con¬‚ict are properly resolved
by appeal to a higher-order principle, indeed a utilitarian principle.
The basic mechanism by which JGT resolves normative con¬‚icts is
its mechanism for balancing the demands of overlapping forms of group
agency, especially DD and VLD. This mechanism must hence be re¬‚ected
in the proper agent-balancing reasons, which in turn should be governed
by analogous but higher-order agent-governing reasons. And the effect of
this is rather ecumenical along the axes now in question. The important
normative status of HMP in JGT provides an analogue to the methods

21 Kamm (1996: 172).
22 Romans 3: 8.

suggested by their partisans for resolution of Ross-style prima facie status
and Sidgwickian vagueness and con¬‚ict. And while the direct consequen-
tialist form of our account to a degree undercuts the intuitive depth of
the difference between positive and negative duties, still we saw in Part
One that MAC in fact undergirds and explains the normative relevance
in certain situations of the difference between omission and commission,
the difference between negative and positive duties, the Pauline Princi-
ple, and the Principle of Permissible Harm. Even the DDE can be given
a similar treatment if our commonsense practice really takes that form.
Actual and normatively weighty forms of group agency have a signi¬cant
normative status according to JGT, and this includes group practices of
distributing risk and harm.
A third sort of structural axis re¬‚ects the degree to which a recon-
struction of commonsense morality includes an account of appropriate
punishment, sanction, or blame. Though there are obvious restrictions
on proper forms of punishment or blame that ¬‚ow from general moral
duties and virtues, some reconstructions give certain forms of blame or
punishment special status, while others do not. Gert, for instance, builds a
concern with punishment into the basic moral attitude toward his moral
rules. Donagan™s reconstruction instead deploys second-order precepts
governing blame. Our concern here is above all with what we should
objectively do and be, not with what people can properly do to us if
we fall short. But still, the mechanism of JGT suggests an intuitive treat-
ment of punishment, noted in Chapter 3. We are enjoined to cooperate
in normatively weighty one-off group acts like that of not murdering
other humans. But if someone violates such a cooperative practice, JGT
implies a change in their normative status. Depending on the existence
of other weighty group actions in which the perpetrator continues to
engage, and the details of our group acts, the propriety of their individual
punishment may then even turn directly on the consequences of such
punishment as evaluated by HMP. In effect, MAC implies a kind of nor-
mative punishment for those who violate certain forms of normatively
weighty cooperation, which is a change in normative status. This may
in turn defeat a normative presumption against a more concrete form of
punishment. This conception of punishment is an intuitive advantage of
the just good theory.
The fourth cluster of structural axes re¬‚ects the fact that reconstructions
differ in the kinds of higher-order principles or other forms of higher-
order structure that they deploy. Still, these various forms of higher-order
structure seem related, so that differences along the fourth set of axes are

mostly more apparent than real. Confucius, the tradition of the Deca-
logue, Buddhism, the Qur™an, Aquinas, Kant, Donagan, and Sidgwick all
recognize a basic role for either a negative or a positive Golden Rule,
an injunction to love others as ourselves, or benevolence. All in this way
suggest a basic normative role for harms and bene¬ts to individuals. And
there are, at least arguably, three general sorts of harms and bene¬ts that
are intuitively relevant to such basic precepts. They are (i) direct harm
or bene¬t to someone, as in giving them goods or taking them away;
(ii) just or unjust relative harm or bene¬t to someone, as in giving them
their just share of goods or less than that; and (iii) respecting or violat-
ing cooperative practices with those individuals. Various forms of more
detailed or even apparently cross-cutting higher-order structure suggested
by some reconstructions in fact track this three-part development of the
traditional general precepts. They provide relevant interpretations of these
three factors. Gert™s reconstruction naturally suggests a division of his ten
moral rules in which the ¬rst ¬ve correspond to a certain interpretation
of (i) and (ii), while the second ¬ve correspond to (iii). The interpretative
element is that all of Gert™s rules grant special salience to harm. Ross™s
reconstruction seems to elaborate each of the three categories in other
ways. The ¬rst category becomes the prima facie duties of benevolence
and malevolence and self-improvement. The second becomes the prima
facie duty of justice. And category three becomes what Ross, somewhat
unusually, calls the category of special obligations, which encompasses the
remaining prima facie duties, which include not only duties of gratitude
and reparation but also the very general duty of promise keeping, incor-
porating the duty not to lie. Sidgwick also recognizes the existence of an
analogous structure within his general virtues of Benevolence and Justice.
Scanlon™s analysis deploys an implicit structure with analogues of (i) and
(ii) in his Chapter 5 and of (iii) in Chapter 7.
JGT suggests a somewhat similar and hence appropriately intuitive
general higher-order structure. Categories (i), (ii), and (iii) correspond
roughly and respectively to Parts Two, Three, and One of this book.
HMP incorporates a conception not merely of the good but of a properly
moralized just good, re¬‚ecting a conception of justice as equality. And
we have MAC with its emphasis on a certain sort of cooperative activity,
and hence another sort of justice. There is spin in this interpretation of
(i) through (iii). But in earlier parts, we saw these various sorts of spin to
be intuitively correct in a general way.
There are other sorts of general structure present in some reconstruc-
tions. We have already discussed the division of duties into duties to self

and others, and into special and general obligations. And of course we
have talked about virtues and duties. But some accounts recognize various
intellectual and practical virtues and duties. These virtues and duties help
to support effective action in accord with the basic moral duties and basic
moral virtues whose organization we discussed in the preceding para-
graphs, or even, in the case of some conceptions of wisdom, incorporate
such moral duties and virtues. MAC naturally implies such supporting
and encompassing virtues and duties, which are at least not determinately
unintuitive, and indeed seem to have been neglected by some reconstruc-
tions simply as a matter of oversight or difference in focus.
We must translate the positions of JGT along our fourth cluster of axes
into the language of accepted reasons. What of the element of higher-
order structure represented by that theory? How can that be incorporated
into PC? The answer is, very simply, as agent-governing reasons (pre-
suming that class to include general rules about the proper balancing
of forms of agency), which govern the proper acceptance of particular
agent-balancing reasons as well as particular agent-constituting reasons.
This constitutes a set of higher-order principles in PC to which partic-
ular decisions can be referred. We will get to its lower-order echo in
agent-constituting reasons quite shortly.
But let me sketch this higher-order component of the translation in
a little more detail. First, there are three principles that constitute the
normative core of JGT. They are HMP, VLD, and DD. They direct us to
weight options for choice in accord with HMP and to act in accord with
that weighting, and yet ¬rst and foremost to act in accord with cooperative
forms of group activity as directed by VLD and DD, though of course
the implications of these tenets rest in turn on HMP. HMP may not tell
us which is the suitably right option if that is less than the best option
from some set of alternatives, but it will tell us which options are better
than others, and that is what is most practically salient in the application
of MAC to moral questions.
Second, there is a little more higher-order structure that we have
already squeezed out of JGT. For one thing, an atomic agent is instructed
to join bene¬cent group acts of which it is not already part on direct
consequentialist grounds, as assessed by appeal to HMP, as long as that is
not otherwise forbidden by the constraints of MAC. For another, TLC,
the three-level conception of the good, extends our sense of when, if
other things are equal according to HMP, one option may be better than
another. It hence extends our weighting of options and also the conditions
relevant to the application of VLD and DD. The respect for these goods

already incorporated in PC involves the acceptance of agent-constituting
reasons of various sorts, at a lower level than is now our concern, but also
of higher-order agent-governing reasons that support this lower-order
content. TLC is incorporated into PC as agent-constituting reasons of
some weight, supported by higher-order agent-governing reasons. These
constitute respect for fundamental agency goods and important tertiary
But the general and for-the-most-part instrumental relationships that
help to constitute genuine tertiary goods and agency goods also provide a
second set of for-the-most-part agent-constituting principles that should
have some weight, though not an indefeasible weight, in PC, and which in
turn seem properly supported by relatively concrete agent-governing rea-
sons of similar weight. These make PC yet more intuitive. Let me explain.
HMP implies in practice that greater risks of greater pains are most nor-
matively weighty. As a matter of general fact, these are caused by acute
physical injuries and constraints, starvation, disease, and death above all,
of animals as well as of humans. So PC should incorporate fairly weighty
agent-constituting reasons that enjoin us to relieve or prevent those things,
as long as that positive action doesn™t involve a worse downside risk. And
PC should re¬‚exively support that with higher-order agent-governing
reasons. But to pursue positive projects of this sort by means that involve
or threaten serious physical injuries or constraints, starvation, disease, or
death almost always involves downside risks that are unlikely to be coun-
tenanced by HMP. So PC should also incorporate fairly weighty agent-
constituting reasons against means that threaten or involve such things,
and second these with agent-governing reasons. Hence there would be
some analogue of priority for principles of nonmale¬cence over benev-
olence in this weighty but defeasible agent-governing layer of PC, and
directed more broadly than the hedonism of HMP suggests on its surface.
There are also at least arguably general and for-the-most-part instrumen-
tal relationships that run in the other direction relative to phenomenal
value. While agency goods or certain tertiary goods are only genuine
when in support of proper projects, and while any general commitment to
respect or favor agency goods or analogous tertiary goods must be hedged
by a quali¬cation that makes the projects in which they are deployed
quite normatively salient, still PC should probably include at least prima
facie reasons to respect and favor such goods, both agent-constituting and
agent-governing. This serves to reinforce and amplify the effects of the
mechanism delivering respect for fundamental agency goods and impor-
tant tertiary goods that are not genuine that I have already discussed,

which involves the existence of bene¬cent group agents extending such
That is more than enough complexity on the fourth axis. Our ¬fth and
sixth axes begin to edge from abstract structure toward speci¬c normative
content. Now a recognizably familiar moral code will begin to emerge.
Scanlon argues that obligations of truth telling and promise keeping are
largely independent of convention, while his antagonists on the ¬fth axis
cede great weight to convention and hence introduce a signi¬cant degree
of relativism into basic duties. The normative signi¬cance of actual nor-
matively weighty group acts according to MAC introduces some element
of conventionality and relativism, though group action does not stretch to
cover all conventions. Still, MAC also recognizes that, other things equal,
truth should always be told in support of proper projects. And MAC
characteristically cedes dominance to the largest group act and hence to
the broadest conventional practice. So on this axis the just good theory
occupies a kind of middle and hence appropriately intuitive ground. We
will return to the details of truth telling in a moment.
A sixth axis of difference that tracks what is at least very close to
being a matter of content rather than structure involves the degree of
revolutionary overthrow of customary forms of life that a code requires.
Some reconstructions of common sense are quite conservative, while
the ancient religious codes apparently require revolutionary psychological
changes that would leave covetousness, unrighteous anger, and even desire
behind us. The revolutionary elements of those ancient codes may be only
¬tfully re¬‚ected in contemporary common sense. But even commonsense
morality may be in some tension with ordinary life. Since we presume
here that our background institutions are appropriately correct, including
those that support and help constitute familiar psychologies, this will not
in the end be relevant to the details of PC. But nevertheless it is clear
that PC should recommend in a general way at least conditional concern
about the propriety of background institutions as evaluated by HMP.
Because of the sensitivity of PC to actual forms of group activity, and
because of its sensitivity to their normative propriety, and given the obvi-
ous facts that forms of group activity may be unstable over time but are
also partly constituted by reasons whose acceptance under local coopera-
tive conditions may be mandated by PC, PC must specify acceptance of
both a ¬rst and more general set of reasons that is relatively independent
of local and current conditions, and a second and more speci¬c set that
is not. Clearly, the agent-governing reasons that re¬‚ect the higher-order
structure in PC and that we have recently surveyed should be re¬‚ected in

otherwise identical agent-constituting reasons of some weight, and this
is at least a crucial part of that component of PC that is independent
of local conditions. But there are also more detailed agent-constituting
and agent-governing reasons, some which are relatively independent of
local context and some of which are not, that are important according to
the just good theory. We will see some effects of this in the next section.
Clearly also, JGT requires that PC specify signi¬cant agent-governing and
agent-constituting reasons to examine background institutions and other
forms of actual cooperation in light of alternative possibilities and HMP,
and to be aware of actual forms of cooperative action and also possible
better alternatives. Such an examination may make revisionary activity
very weighty according to JGT. Still, as I said, our general presumption
in this chapter that the background conditions of our lives are appropriate
makes these ultimately very important issues largely irrelevant for now.


The seventh, last, and very large set of axes on which reconstructions of
commonsense morality differ concern relative details of normative con-
tent. The location of PC on these relatively concrete axes provide it with
a recognizable and appropriately intuitive shape. Aside from the familiar
higher-order structure of PC, the guts of it are found here.
Differences on these axes are differences about the relatively detailed
form of agent-constituting and agent-governing reasons, and sometimes
also agent-balancing reasons, to be properly incorporated into a code.
Because we must compare and contrast content aspects of both duty-based
and virtue-based reconstructions that deploy different sorts of higher-
order organization, I will organize our review by the particular set of
categories they occupy according to the framework of JGT.
Benevolence and justice are already re¬‚ected in the higher-order struc-
ture of PC. But various reconstructions of commonsense morality deploy
various other individual virtues that help to constitute successful con-
tinuing agency. Temperance, fortitude, prudence, courage, good tem-
per, moderate ambition and pride, Confucian chung (doing one™s best),
patience, and a lack of servility are one set, which provide affective con-
ditions of effective agency. I will call these agency virtues. There is also
the intellectual group of intelligence, Buddhist right views, and practical
wisdom. The latter incorporates the Confucian wisdom of knowing oth-
ers™ hearts and Kantian knowledge of one™s own motives. These are also
arguably agency virtues for the most part. There is also a traditional set

of socially attractive virtues, like Aristotelian friendliness and ready wit


. 9
( 10)