. 24
( 38 .)


Brute Rationality

taken as giving the meaning of ˜irrational action.™ Rather, it will be argued
that the meaning of ˜irrational action™ is better understood as given by a
substantive reference-¬xing de¬nition “ that is, by a description of those
substantive characteristics in virtue of which an action would prompt the
relevant attitude in the overwhelming majority of people. It is, of course, an
open question whether the substantive de¬nition is suf¬ciently accurate “
that is, whether it manages to capture, with an allowance for vagueness,
those actions that people would call ˜irrational™ if it were explained to them
what this term was meant to capture: a type of action that no one could
sincerely recommend. But the potential for error here does not matter a
great deal, unless the divergence is signi¬cant. This said, the account of
regarding an action as irrational is as follows:
A1 A person regards an action as irrational iff that person cannot
see, and does not believe that there are, any consequences
of the action that could allow someone sincerely to advise
someone else to do it.3
With this account, it is possible to provide an account of objective irra-
tionality in the following way.
A2 An action is objectively irrational iff virtually everyone
would regard the action as irrational, if they were fully
informed about all nontrivial consequences of the action.
I do not mean to beg questions by using the word ˜consequences™ rather
than ˜features™ in A1 and A2. I use ˜consequences™ for purposes of clarity,
because all the features that appear in A3 below are consequences. Should
this turn out to be a mistake, I do not believe anything of importance in
the rest of the account is affected. Moreover, I grant that there is some
plausibility to the idea that there are moral rules “ for example, ˜Keep
your promise™ “ that can sometimes provide rational justi¬cation without
the prospect of any of the right sort of consequences for anyone. But
the applicability of such a moral rule would, in my view, only provide a
rationally justifying reason, even if it provided a moral requirement.4
3 This attitude is not to be taken in an overly intellectualist way, so that a person would
have to think to himself that he cannot see any consequences that would allow for a sincere
recommendation. It is enough that the person would be puzzled if someone were to try
sincerely to recommend it “ or even allow it “ based on the likelihood of the various
possible consequences.
4 I do not believe that any problems of circularity would result if my account of ratio-
nality were to be used in the relevant account of morality. One way of avoiding such a

Two concepts of rationality

As was mentioned above, A2 should not be taken as suggesting that
to call an action ˜irrational™ is to assert that there is a nearly unanimous
agreement in judgments of a certain sort, any more than to call a banana
˜yellow™ is to make such a claim. But it is not irrelevant to the mean-
ing of the word ˜irrational™ that there is such near unanimity. For this
near unanimity, because it allows for ostensive teaching of the word, also
allows it to share the grammar of objective descriptive words, and to have
an objective referent. Making use of this fact, and substituting a plau-
sible extensional equivalent for the right-hand side of A2, we get the
A3 An action is objectively irrational iff it involves a nontriv-
ial risk, to the agent, of nontrivial pain, disability, loss of
pleasure, or loss of freedom, or premature death without a
suf¬cient chance that someone (not necessarily the agent)
will avoid one of these same consequences, or will get plea-
sure, ability, or freedom, to a compensating degree.5,6
One signi¬cant misunderstanding that may arise at this point comes
from reading the word ˜irrational™ in A2 independently of the technical
and syncategorematic phrase ˜regard as irrational,™ as it is de¬ned in A1. If
one reads A2 in this mistaken way then A3 is unlikely to seem as plausible
problem would be by explicitly restricting the notion of rationality to the consequential-
ist version when developing a moral theory. Another more interesting and perhaps more
plausible way would be by solving a sort of ˜normative differential equation,™ the result of
which would be the limit of the following series. The ¬rst term is the consequentialist
account of rationality, the second term is an account modi¬ed by moral reasons yielded by
the moral rules that emerge from a moral theory that takes rationality to be as given by the
¬rst term, the third term is an account modi¬ed by the moral reasons yielded by the moral
rules that emerge from a moral theory that takes rationality to be as given by the second
term, and so on. There is no reason to doubt that such an account would fail to yield
determinate accounts of both rationality and morality. This is because the nonconsequen-
tialist reasons will always be of relatively small importance compared to the consequentialist
ones, as long as morality, as a system, is rightly regarded as primarily concerned with the
welfare of sentient beings.
5 My debt to Bernard Gert is very great here. See B. Gert (1998), chs. 2 and 3. At the time
I incurred this debt, Gert did not make the distinction between objective and subjective
rationality, or formally distinguish the justifying and requiring roles of reasons. He has since
modi¬ed his view in response to the arguments presented here.
6 The phrases ˜a suf¬cient chance™ and ˜to a compensating degree™ of course retain a normative
aspect. These normative phrases could be eliminated by specifying, descriptively, the types
of trade-offs that the overwhelming majority of people regard as rational. I have not
eliminated them because to do so would make A3 too cumbersome to serve its illustrative
purpose. For this same reason, A3 does not re¬‚ect the possibility that egregious harms to
others may sometimes make an action irrational.

Brute Rationality

as it is here claimed to be. For there is a signi¬cant group of philosophers
who, if fully informed, would nevertheless regard it as perfectly rational “
in a sense different from that given in A1 “ for an agent to chop off her little
¬nger, given that the agent had a sincere desire to do so, and no oppos-
ing desires of suf¬cient strength. Humeans such as Bernard Williams hold
such a view.7 Now, there are a whole range of arguments against views
that invest desires with this kind of normative signi¬cance. For example,
Joseph Raz, Jonathan Dancy, Thomas Scanlon, and Warren Quinn have all
recently argued that it is not desires that have normative relevance, but the
reasons behind them, and that if there are no reasons behind one™s desires
then those desires are whims at best and pathological at worst.8 These
arguments are very persuasive, and they certainly support the general view
offered here. But in fact they are not relevant at this point. Rather, what
is important is to recall that we arrived at A1 by analysis of what we mean
when we use terms like ˜irrational™ in a certain fundamental way: the way
indicated by (1). The present account relies on the supposition, defended
below, that there is suf¬cient agreement in what people regard as irrational
(in the relevant sense) to yield an objective matter of fact as to what kinds
of actions really are irrational “ in the sense of ˜objectively irrational™ “
and what kinds are not. When ˜regard as irrational™ is understood in this
way, A3 is extremely plausible. Additional strong support for A3 will come
from discussion of subjective irrationality: the kind of irrationality that has
a much closer connection to practical mental functioning, moral respon-
sibility, free will, and so on. For there will be a simple account, in terms of
A3, of subjective irrationality. The plausibility of this account will provide
additional support for A3.
As we saw in chapter 1, any account of subjective rationality should be
consistent with the idea that a fully informed agent, performing an objec-
tively rational action, will typically be performing a subjectively rational
one also.9 Call this the ˜objective-subjective implication.™ Against A3, and
making use of this implication, many theorists have claimed that all actions
that harm others for morally insuf¬cient reasons are not only immoral,
but are also irrational; in other words, that moral requirements are also

7 See Williams (1981), p. 105.
8 Quinn (1995), p. 195; Dancy (2000), pp. 35“38; Raz (1999b), pp. 50“64; Scanlon (1998),
pp. 35“42.
9 More accurately, such an agent must at least have the possibility of thereby performing
a subjectively rational action: such an action should only be subjectively irrational if it is
done ˜for the wrong reasons.™ This quali¬cation will not be important in what follows.

Two concepts of rationality

rational requirements. A3, taken together with the objective-subjective
implication, certainly is inconsistent with this claim, for it allows it to be
subjectively rational to hurt other people for pro¬t. Ultimately the best
argument against this view will be the greater adequacy of the view offered
here, taken as a whole.10 But more pointedly, it can also be urged that this
objection simply goes against the way in which the notion of ˜irrationality™
is understood by competent language speakers. Of course I do not mean
to appeal to intuitions about the use of the very word ˜irrational,™ much
less to the phrase ˜subjectively irrational.™ The ¬rst of these is rarely used by
normal people, and the second is a technical term. Rather, I mean that it
is very plausible that there is overwhelming agreement if it is understood
that we are using the term ˜subjectively irrational™ to categorize those
actions that involve a kind of failure in practical mental functioning that
is relevant to questions of moral responsibility and so on, whether or not
those failures are suf¬ciently extreme to have much importance in partic-
ular cases. That is, ˜subjectively irrational™ is meant to collect the spectrum
of actions that range from ˜silly™ and ˜stupid,™ through ˜boneheaded™ and
˜a bad idea,™ all the way up to ˜crazy,™ ˜insane,™ and worse. We generally
would not say, for example, that it is irrational, in this sense, to embezzle
money, or to cheat on one™s partner, or even directly to hurt others by
acting on vengeful impulses. Or, if we do say that such actions are irra-
tional, the reason we offer is almost always ˜because you might get caught,™
and almost never ˜because it might hurt someone.™ In general we only say
that an immoral action is irrational in virtue of the harm it does to others, if
that harm is very signi¬cant. For example, we might well say this if we
discovered that our friend had someone tied up in the basement, and was
preparing to cut off that person™s toes merely as an experiment. But though
these types of actions receive a disproportionate amount of attention in
philosophical and ¬ctional literature, they are far from the typical cases of
immoral action. In typical cases it is the applicability of a reason like the
following “ that one™s friend may get caught and punished “ that explains
why one feels the need to cite some justifying reason in recommending
an immoral action to anyone. But such a reason need not be of the right
kind to morally justify the action. Of course, the rational permissibility of a
great deal of petty immoral action does nothing to argue against the view
that morally required action is also, and always, rationally permissible. A3,

10 See also Svavarsd´ ttir (1999), pp. 161“219. Svavarsd´ ttir makes use of many of the same
o o
intuitions about the rationality of action that this book has relied upon.

Brute Rationality

and the associated account of subjective rationality, does not speak in any
way against the rationality of acting morally; it only speaks against the
necessary irrationality of acting immorally.
A more plausible but equally ¬‚awed argument that moral duties provide
altruistic rational requirements comes from considering possible harms to
one™s family or friends as a result of one™s actions. Suppose, for example,
that someone with a fairly high-paying job and a good deal of savings
decides to give away virtually all of his money, to quit his job, and to
get a new job at a very low salary, working for a nonpro¬t organization
that promotes the availability of contraception and prenatal care in Third
World countries. Though this behavior sounds extreme, it does not sound
irrational until we add the fact that the man has a wife and children towards
whom he has speci¬c duties, and who will suffer a great deal as a result
of his ˜altruism.™ This might seem to suggest that speci¬c duties to others
can provide altruistic reasons of substantial (rationally) requiring power.
But this example is not as simple as it looks. First, it may well be that
in some cases the best explanation for this sort of behavior is that the
agent is suffering from a mental illness: that may be the best explanation
for many drastic changes in behavior. In such cases the action may be
irrational, in the sense of proceeding from a failure of practical mental functioning.
But not all such cases need be explained in this way, and in any case we
are not considering this sense of ˜irrational™ at the moment. Moreover,
even the suspicion that such action is irrational in this latter sense depends
covertly on the assumption that the agent cares a great deal about his
wife and children. If we explicitly reject this assumption, and are really
convinced that a father has grown so estranged from his wife and children
that his duties to them, though real, are felt by him to be rather a burden,
then his extreme altruistic behavior towards strangers, described above,
no longer seems irrational even in the ˜mental functioning™ sense. It only
seems to be tinted with an uncommon degree of coldness and cruelty.
If we take care to separate questions of proper mental functioning from
questions of what actions can be sincerely recommended, then ˜cruel-
altruistic™ action no longer seems irrational in the fundamental sense at
issue here. For the question of relevance is: could someone concerned
for the agent sincerely recommend to him that he allow his own family
to suffer for the sake of nameless strangers? And the answer is ˜Yes.™ Peter
Singer, for example, has recommended exactly this.11 And even the people

11 Singer (1972), pp. 229“42.

Two concepts of rationality

who do not act as he suggests are not puzzled when they consider the
basis on which Singer makes his recommendation. That is, people do
not regard such action as irrational, in the sense of regard-as-irrational
given in A1.
It is important to note here that no argument is being offered here for the
correctness of the substantive description of those actions that one can sin-
cerely recommend. The above discussion is only attempting to describe the
wide range of actions which, on consideration, we think people can and
cannot recommend in this way. The current aim is to get the extensions of
the concepts ˜objectively rational action™ and ˜objectively irrational action™
right. One of the main features of the description is that the existence of
compensating (or even more than compensating) bene¬ts for others does
not generally prevent us from making comprehensible recommendations
of action in the face of the reasons those bene¬ts provide. No argument is
being offered for this description, because if it is correct, there will not be
any argument possible. To what normative principle could such an argu-
ment possibly appeal? Any principle that was not exactly equivalent to the
correct description would be wrong. But any principle that was exactly
equivalent would be exactly equivalent, and therefore could not play a jus-
tifying role in an argument for the correctness of the description: it would
just be a repetition of the description. When we are making accounts of
rationality as the fundamental normative term, a correct description of the
fundamental way we should always behave is the end of the normative road.
It should not surprise us that this is the correct method for fundamental
normative theory, even if it is not the correct method for, say, moral theory.
For there cannot be any more fundamental normative principles that could
ever provide a normative argument in favor of fundamental normative prin-
ciples. On the other hand, moral theory generally does appeal to more
fundamental notions. At bottom, ˜objectively rational action™ must sim-
ply be described, based primarily upon the way normal language speakers
learn to use the relevant normative words like ˜reason,™ ˜makes sense,™ and
˜recommend.™ If someone acknowledges that, given the meanings of these
terms, there is a reason against performing an action “ a reason of the
sort that it makes no sense to act against without some countervailing
reason “ and that there is no countervailing reason that anyone could sin-
cerely recommend acting on (so that, at least in that person™s estimation,
the action is objectively irrational), then she cannot ask ˜But is it really
true that there is a reason against it, and that it would make no sense to
do it?™

Brute Rationality

The above remarks might be summarized as two claims:
(a) there are no untaught-but-intuitively-accessible meanings for norma-
tive words like ˜reason,™ ˜justify,™ or ˜ought™ that make it possible to ask
whether it is really true that the things that everyone takes to provide


. 24
( 38 .)