. 27
( 38 .)


and acts depends instead on what we believe, or given the evidence, ought
rationally to believe.”29 It is to Par¬t™s credit that he notices the relevance,
to subjective rationality, both of actual beliefs, and of beliefs that we ought to
have. But he does not note that what we ˜ought rationally to believe™ can
only make actions rationally required, and cannot ever provide a rational
justi¬cation unless we also believe it. That is, he does not consider the sort
of cases that provide counterexamples to R2. Even if Par¬t did make this
distinction, his account of the relation between objective and subjective
rationality would only be as adequate as R3, for it does not mention

28 Philippa Foot has noticed something like this, in recognizing the different logical roles
played by the objective interests of the agent, and the objects of the subjective desires of the
agent. See Foot (1978b), p. 156. And David Copp makes a similar distinction between the
objective needs and the subjective values of agents in determining the rationality of their
actions. See Copp (1995), pp. 172“85. My view explains these asymmetries by reference
to an agent/other asymmetry in the notion of objective rationality, which has itself been
independently motivated. It should also be noted that neither Foot nor Copp make any
explicit distinction between objective and subjective rationality.
29 Par¬t (1997), p. 99.

Two concepts of rationality

the differential relevance of the agent™s motivation to the justifying and
requiring roles of reasons.
The above considerations lead to a ¬nal and most extravagantly ad hoc
R4 An action is irrational if, relative to the beliefs that the
agent should have about harms the agent may suffer, and to
the beliefs about bene¬ts to others that the agent both (a)
actually has and (b) is moved by, it is objectively irrational.30
One obvious problem with this de¬nition is that beliefs motivate people to
different degrees. This problem, though it provides a strong objection to
the de¬nition, could be overcome with a de¬nition still more complicated.
But that does not matter. The point to note here is that de¬nition R4 is
so ad hoc that, even if it were correct, it would fail to supply almost any
insight into the nature of irrational action. A modi¬ed de¬nition would
fail even more spectacularly. The de¬nition does nothing to suggest a uni-
fying principle behind each of the counterexamples. The lesson we should
learn from the failure of de¬nitions R1 through R4 is that the relation
between objective and subjective rationality is not fruitfully conceived as
these de¬nitions have conceived it. That is, subjective rationality is not
fruitfully conceived as objective rationality relative to some special class of
One of the reasons for the failure of the above de¬nitions should be
obvious from the nature of the counterexamples to R1 and R2. The
counterexample to R1 took advantage of the fact that reasons with con-
siderable requiring strength can be relevant to the rationality of an agent™s
action even though the agent does not see those reasons: it is suf¬cient that
the agent should see them. The counterexample to R2 took advantage of
the fact that reasons that primarily play a justifying role are not relevant
to the rationality of an agent™s action unless they are actually seen by the
agent. R3 implicitly took account of the justifying/requiring distinction by
treating primarily justifying reasons (i.e., those involving bene¬ts to others)
differently from reasons that also require (i.e., those involving harm to the
agent). This is why it represented such an increase in adequacy. But it also
failed, for the justifying power of a reason is not relevant to the subjective
rationality of an action unless it provides the agent with a motive. Unless
30 This de¬nition is not meant to involve any controversial claims about beliefs being moti-
vating on their own. As far as R4 is concerned, a belief can move a person in virtue of
an antecedent desire, or by some other means.

Brute Rationality

one has the justifying/requiring distinction in place in one™s account of
objective rationality, it will be almost impossible to understand these fail-
ures to de¬ne subjective rationality in terms of objective rationality. For
there does not appear to be any other explanation for the fact that some
considerations continue to be relevant to the rationality of action regardless
of whether they are actually believed by the agent, while other consider-
ations do not appear to be relevant unless they are actually believed. The
difference here is emphatically not the difference between culpable and
nonculpable ignorance.31 For even if we eliminate all nonculpable igno-
rance by ¬at, stipulating that the agent should be aware of all the relevant
information, none of the counterexamples to R1 through R4 are altered
in any way. For none of those counterexamples made any reference to
nonculpable ignorance.
Given the above remarks, the reader may now be expecting a de¬nition
of irrationality that explicitly mentions the requiring and justifying roles
of reasons.32 But what we would really like is an account that explains why
belief and desire have a differential impact on the requiring and justifying
roles of reasons, when the move is made from objective to subjective
rationality. This would be superior to an account that merely took account
of this differential impact. Here then is the of¬cial de¬nition of subjectively
irrational action:
R5 An action is subjectively irrational iff it proceeds from a state
of the agent that (a) normally puts an agent at increased risk
of performing objectively irrational actions, and (b) has its
adverse effect by in¬‚uencing the formation of intentions in
the light of sensory evidence and beliefs.33
R5 explains the following facts. It explains why it is relevant to the ratio-
nality of an action that an agent should have a belief that his action will
cause him harm, and not exclusively relevant that he actually does have
such a belief. For if an agent fails to have such a belief when, in the light
of sensory evidence and other beliefs, he should have had it, then he has
31 Although this distinction is, of course, relevant to the rationality of action in other ways.
32 Indeed, Bernard Gert has recently been pushed by the preceding criticisms into just such
a view.
33 ˜Normally,™ because science-¬ction worlds in which a guardian angel secretly guarantees
that all my actions are objectively rational are nevertheless worlds in which I can act
irrationally. A useful gloss of ˜normally™ is ˜in the circumstances responsible for the devel-
opment of the concept of subjectively irrational action.™ In this connection, see Millikan
(2000), pp. 61“68.

Two concepts of rationality

failed to see something that he should have seen, and, given that what he
failed to see is a requiring reason, his failure shows that he is at increased
risk of doing things that are objectively irrational. This explains the coun-
terexample to R1. On the other hand, R5 also explains why it is relevant
to the rationality of an action that an agent actually have a belief about the
compensating bene¬ts of his action and other primarily justifying reasons,
rather than merely that he should have such a belief. For if the agent merely
should have such a belief, but does not have it, and acts in a way that he
knows will bring him harm, this shows that he is insuf¬ciently averse to
the perceived harm “ a requiring reason “ and thus that he is at increased
risk of doing objectively irrational things. This explains the counterexam-
ple to R2. A similar explanation accounts for the fact that beliefs about
the bene¬ts to others must supply a motive for the agent if they are to
rationally justify her action. That is, suppose one knows two things about
one™s action: that it will cause one a lot of pain, and that it will prevent
someone else from suffering a comparable amount of pain. If one does
not care that the action will prevent someone else from suffering pain, and
goes ahead and does the action, then one™s action involves a failure to be
appropriately averse to one™s own pain. This explains the counterexample
to R3.
Clause (b) links subjective irrationality explicitly to the will. It thus
eliminates factors such as blindness and clumsiness as relevant sources of
increased risk of acting in an objectively irrational way. If someone is blind
or clumsy, and consequently at increased risk of harming himself, we do
not call his actions ˜irrational,™ but ˜unfortunate.™34 Clause (b) also provides
some explanation for the fact that ˜irrational™ is a term of practical criticism.
For while criticism of actions that stem from blindness or clumsiness would
typically have no point, actions that meet clause (b) are often ones that
could be avoided by an effort of will, and are consequently ones for which
criticism could have a point. Indeed, part of the point of having the ˜mental
functioning™ concept of irrational action (less metaphorically: part of what
explains the presence in the language of words that allow one to have this
concept) is that it helps us to provide the very criticism to which such
actions are liable. In some cases, of course, the state of the agent from
which the action proceeds is a disability of the will “ say, an addiction “
so that criticism may be doomed to failure. Thus, while ˜rationally ought™

34 This remark indicates the sense of ˜increased™ in ˜increased risk.™ It means ˜greater risk
than the population at large.™

Brute Rationality

generally implies ˜can,™ it does not always do so. Addictions, phobias, and
compulsions certainly cause people to do things that, rationally speaking,
they ought not do. And yet it is unclear whether, in any important sense of
˜can™ that goes beyond mere physical ability, they can always avoid doing
those things. Clause (b) also suggests that some actions stemming from
extremely low intelligence should be classi¬ed as irrational. Although this
may go against common usage to some degree, it is worth noting that
extremely low intelligence is linked to diminished moral responsibility,
diminished competence to give consent, and even diminished freedom of
the will, in the same way that more ˜intelligent™ irrationality is. Moreover,
the sorts of everyday irrational actions with which this book is concerned
are often called ˜stupid™ or ˜dumb™ even when they do not proceed from
low intelligence at all. That these words are nevertheless used suggests that
actions that stem from actual stupidity share important features with more
paradigmatically irrational actions.35
On this account of practical rationality, objective rationality may be seen
as playing a role similar to that played by truth in accounts of theoretical
rationality.36 That is, one might de¬ne ˜irrational belief™ in the following
B1 A belief is irrational iff it proceeds from a state of the agent
of a kind that (a) normally puts an agent at increased risk
of believing false things, and (b) has its adverse effect by
in¬‚uencing the formation of beliefs in the light of sensory
evidence and other beliefs.
Although it is beyond the scope of this book argue for it, B1 does seem to
capture much of the content of ˜irrational belief,™ where irrational beliefs
are ones that we take to re¬‚ect negatively on the cognitive capacities
of the believer. The plausibility of B1 as an account of irrational belief
would reinforce this book™s account of the relation between objective and
subjective rationality. But of course there are so many differences between
beliefs and actions that we should not expect B1 to be perfectly adequate,
even if R5 is correct.
R5, taken together with A2 and A3, accounts for the plausibility of a
great diversity of alternate accounts of rationality, and for our intuitions
in some puzzling cases. For example, R5 accounts for the plausibility of

35 I thank Maria Victoria Costa for these points.
36 In this connection, see Anscombe (1995), pp. 32“33.

Two concepts of rationality

almost all purely formal accounts of rationality, such as ˜considered pref-
erence™ accounts. For if one is acting against one™s considered preferences,
this indicates a sort of malfunctioning in the will that places one at increased
risk of doing objectively irrational actions. But R5 also accounts for the
possibility of irrationally acting in accord with one™s considered preferences,
in those cases in which those preferences are themselves objectively irra-
tional. Purely formal accounts of practical rationality notoriously fail to
do this, so that it becomes perfectly in accord with rationality to prefer
the destruction of the entire world to the scratching of one™s little ¬n-
ger, or to opt, for no reason, against taking the medicine that will return
one to perfect health and a pleasant life.37 R5 also accounts for the nor-
mativity of instrumental reason, since failures in instrumental rationality
are certainly characteristic of a state of increased risk of doing objectively
irrational actions. This is true even if the particular failure of instrumental
rationality turns out to be ˜for the best.™ For suppose that it was blind
luck that the failure was for the best. Then similar failures in the future
are likely to be harmful. On the other hand, suppose that the failure in
instrumental rationality was for the best because the end to which one
failed to take appropriate means was a harmful end. In this case, even if
one is ˜cured™ of having the harmful end, one still has a problem, and it is
from this problem that the original failure proceeded. Similarly, R5 explains
why it counts as irrational to act against one™s own normative judgments “
at least if the capacity for coming to such judgments can plausibly be seen
as a tool for increasing one™s chances of acting in an objectively rational
way. Thus Thomas Scanlon™s restricted understanding of ˜irrational™ can be
understood as corresponding to one speci¬c form of a much more general
phenomenon, and its negative normative signi¬cance can be explained. R5
also accounts for the normative signi¬cance of formal restrictions on pref-
erences, such as transitivity, without making it irrational to change one™s
preferences, and without falling prey to an objection that affects ˜dutch
book™ arguments for the same formal features: the objection that one of
one™s strong preferences may be to fall victim to the dutch book. In all
these cases “ failure to act on one™s considered preferences, failure to have
objectively rational ends, failure to take the proper means to one™s ends,
failure to act on one™s judgment regarding how one ought to act, failure to
be suf¬ciently consistent in one™s preference-ranking “ there is a different
type of state that explains why one is at increased risk of doing objectively

37 Hume (1978), p. 415; Williams (1981), p. 105.

Brute Rationality

irrational actions. This shows how R5 might be used to create a sort of
taxonomy of practical irrationality. In connection with the idea of such a
taxonomy, it is worth remarking that there is a sort of prejudice against
substantive accounts of rationality that offer a list of reason-providing con-
siderations, such as pain, premature death, knowledge, ability, and so on.38
Such lists are accused of being arbitrary. But it is amazing how many formal
accounts of rationality offer similar lists without incurring the same sort of
criticism.39 True, these lists are lists of formal conditions, which are more
appealing to philosophers. But all that the members of these lists of formal
conditions typically have to recommend themselves is a very high degree
of surface plausibility “ just a bit less surface plausibility, in my estimation,
than the claim that premature death is bad, and pleasure good. R5 offers
a way of justifying these lists of formal conditions.
R5 also accounts for the subjective irrationality of behaviors that result
from mental disorders such as schizophrenia, which other accounts will
have a surprisingly dif¬cult time explaining. For example, suppose that
Jim hears voices in his head telling him that wearing tin foil under his
clothes is the only way to protect himself from some dangerous rays. Let
us grant, for the sake of argument, that wearing tin foil in this way is suf¬-
ciently uncomfortable that one requires a justi¬cation for doing so. Now,
if Jim knows nothing of psychology, it is not completely unreasonable “
in a certain sense “ for him to take the voices in his head that identify
themselves as powerful aliens to be powerful aliens. And it would not be
unreasonable therefore for Jim to believe that the tin foil they recommend
really will help him avoid the effects of the dangerous rays. But despite
all this ˜reasonableness,™ someone like Jim, who hears voices telling him
to wear tin foil under his clothes and who therefore wears it, is acting
irrationally. The Noah of the Bible, on the other hand, was not acting
irrationally in suffering the ridicule of his neighbors by building the ark.
Epistemically, Jim and Noah are in pretty much the same situation. How
then, can we call the one™s actions ˜irrational™ and the other™s ˜rational™?
By the appeal that R5 makes to the risk-increasing nature of the state that
causes the relevant behavior.


. 27
( 38 .)