<<

. 20
( 38 .)



>>


Words such as ˜would,™ ˜most,™ and ˜best™ can also help to import the
assumption. For example, Michael Smith suggests a way in which facts
about what we ought to do, all things considered (in my terminology,
what is objectively rational) are ¬xed by the desires we would have if fully
rational. His suggestion is that what we have all things considered reason
to do is:
¬xed by the relative strengths of these [fully rational] desires: that is, by facts
about what our fully rational selves would most want us to do in the relevant
circumstances.7

That is, Smith equates the normative strengths of reasons with the unique
strengths of certain hypothetical desires.
Critics of ideal motive accounts also make this assumption on behalf of
the theorists they criticize. Connie Rosati, for example, in her criticism
of ideal motive accounts of the good, rightly points out that the different
ways in which one might become fully informed are likely to result in one™s
ending up with different motives. From this she concludes that, when one
is constructing a full-information ideal motive account of the good:
the process of fully informing a person will need in some way to offset the effects
of experiential ordering, by requiring, for instance, that a person experience all
lives . . . in all possible orders.8

6 Darwall (1990), p. 262, italics mine. Although he does not state it explicitly, it is clear that
Darwall makes the uniqueness assumption for nonnaturalistic versions also. See also Smith
(1996), p. 167 and Brandt (1979), p. 15.
7 Smith (1996), p. 167, emphasis added. See also Sidgwick (1981), pp. 111“12; Rawls (1971),
pp. 417“18; Brandt (1979), pp. 126“29; Railton (1986), p. 16; Pettit and Smith (1993),
pp. 53“79; Rosati (1995), p. 302, esp. n. 15.
8 Rosati (1995), p. 309. It is true that Rosati is concerned with accounts of what is good for
a person, and not with accounts of normative reasons. But the example illustrates the same
assumption about the uniqueness of ideal motivation. Moreover, an account of the good
for a person can plausibly be regarded as yielding at least a partial account of that person™s
reasons for action.

115
Brute Rationality

But why will these ordering effects need to be offset? The answer seems
to be: ˜Ideal motive theorists will need to offset these effects, because for
their view to be viable, there must be a unique ¬xed set of motivations that
they can claim the ideal agent will possess.™ But why is this? This question
will be especially hard for Rosati to answer, since she concedes that the
effects of experiential ordering need not be regarded as the result of any
cognitive problem.
It may well be that some ideal motive theorists use de¬nite descriptions
such as ˜the motivational pull a consideration would generate under ideal
circumstances™ without committing themselves to the uniqueness assump-
tion. They may be using the phrase ˜the motivational pull™ in the same
way one uses the phrase ˜the height of a grown man.™ No one thinks
that there is a unique height shared by all grown men. Rather, it is clear
that this latter phrase indicates a range of heights. Moreover, while the
boundaries of the range may be vague, the whole range is not merely
the result of vagueness. Both 5 8 and 6 1 clearly count as ˜the height
of a grown man.™ Most ideal motive theorists do not appear to be using
the phrase ˜the motivational pull™ in this looser way.9 But whether or
not they are, and whether or not there are other theorists who would
immediately accept this chapter™s proposed modi¬cation, no ideal motive
theorist has explicitly denied the uniqueness assumption, much less explored the
theoretical consequences of such a rejection. This chapter should therefore be
of use even to those who have no prior commitment to the uniqueness
assumption.
This chapter™s objection to the uniqueness assumption is not that moti-
vation can only be measured with a certain roughness, and that more than
one estimate of the strength of a reason might therefore be acceptable.
Of course this is true, especially for hypothetical motivation. But it is also

9 For example, Smith (1995), pp. 118“25 believes that the desires of rational agents will
converge. Connie Rosati, for the reasons given above, also seems committed to the need
for ideal motive theorists to make the assumption, although she is not herself such a theorist.
Stephen Darwall almost always writes as if a given consideration either would, or would
not, motivate a fully rational agent. See, e.g., Darwall (1983), pp. 134“38. But Darwall is
also one of the few theorists who have explicitly recognized the possibility that there might
be ˜no single correct answer to the question of what reasons there are for a given person to
act in a given situation.™ See Darwall (1983), p. 241. In fact, however, this claim by Darwall
is a consequence of his view that there may be no single correct theory of decision under
uncertainty. The absence of such a single correct theory is irrelevant to the question of
whether there is a single correct answer to the question of how strong a particular reason
is. Rather, it results only in an indeterminacy in what an agent has all-things-considered
reason to do.

116
The relation to other views

likely that many ideal motive theorists would tolerate the small range of
acceptable strength values that such vagueness implies. Rather, the objec-
tion is that there may be a range of clearly acceptable degrees of motivation
that a given consideration could generate, even in an ideal agent. The
existence of such a range is different from the existence of vagueness. If
two people are told to arrive at a party between 8:00 and 10:00, and one
arrives at 8:30 and the other at 9:15, then the reason that both are immune
from the charge of being too early or too late is not that there was a certain
ineliminable vagueness in the details of the invitation. Rather, both have
arrived squarely in the middle of a range of times that are all, in point of
being on time, equally ˜ideal.™ And the speci¬cation of this range requires
two values.
It is extremely easy to misunderstand the position that results from
the rejection of the uniqueness assumption. One might think that the
suggested range of acceptable degrees of motivation in an ideal agent
will correspond somehow to a range of acceptable degrees of normative
strength. But this is a mistaken interpretation. In fact, it is not even clear
that it is a coherent interpretation. For what could it mean that a speci¬c
reason had a range of normative strengths? One might take it to mean that
the reason could have different normative strengths on different occasions.
What would this strength depend on? The obvious suggestion is: the actual
motivational strength that it provides on each occasion. But this merely
confuses normative strength with motivational strength.10 Nor can one
associate the normative strength on a speci¬c occasion with the speci¬c
motivational strength that the reason would provide to the agent under
ideal circumstances. For we are trying to clarify the position that results
from the rejection of the uniqueness assumption. And the rejection of the
uniqueness assumption is simply the denial that there is any such speci¬c
motivational strength.
The correct interpretation of the rejection of the uniqueness assumption
is the following. The unique value that was called ˜normative strength™ is
replaced by two values. These two values cannot sensibly be called ˜min-
imum normative strength™ and ˜maximum normative strength.™ Rather,
they might be called ˜minimum rationally permissible degree of motiva-
tion™ and ˜maximum rationally permissible degree of motivation.™ When
an agent is faced with a decision, the rational options open to him will
include all the actions that could result from any combination of rationally
10 Relatedly, it also confuses the question of the existence of a reason, with the question of
whether or not an agent bases his or her action on that reason.

117
Brute Rationality

permissible degrees of motivation “ regardless of how likely it is that this
particular agent will choose them, given his actual motivational setup.11
The agent could do any of these actions, and be acting rationally. Of course,
the agent™s action must, in some intuitive way, be based on the reasons that
make it fall into the set of rational actions, if that agent is to be counted as
acting rationally, and if the action is to be regarded as subjectively rational.
And basing an action on a reason may well involve being motivated by that
reason. But the question of the existence of reasons, and of their strengths “
and therefore the question of the objective rationality of an action “ is prior
to the question of whether or not an agent bases his action on the reasons
that make it a rational option. This is also true for accounts that accept the
uniqueness assumption, as the following example illustrates. Suppose that
the uniquely rational thing for an agent to do in a certain situation is to
take some medication in order to save her life. But suppose that the agent
takes the medication only because she hates to see pills lying around. This
would not make it any the less true that there was a strong reason to take
the pills, and that because of this reason the rational thing to do was to
take the pills.
By allowing that there might be a range of rationally acceptable options
that are independent of (but, barring irrationality, include) what the agent
is likely to do, it becomes possible to make the following common sort
of claim: it would be perfectly rational for Bill Gates to sell off all his
shares of Microsoft, donate the proceeds to relieve the suffering caused
by the hurricane in Honduras (and a mass of other suffering), and begin
to work on salary for someone else.12 The fact that someone is very
unlikely to perform a certain action does not make it impossible for us
to make normative judgments about that action.13 Perhaps it is true that
11 The argument of this chapter therefore suggests an internalist position between those
which Michael Smith has called ˜Humean™ and ˜Kantian.™ For it allows that the normative
capacities of reasons can be characterized in an objective way (as the Kantian internalist
maintains), and it also allows that different agents might be motivated to differing degrees
by the same reason (as the Humean internalist maintains). See Smith (1995), p. 118.
Smith™s false assumption that Humean and Kantian internalism exhaust the ¬eld of possible
internalist views is the result of his uncritical acceptance of the uniqueness assumption,
and the association of normative strength with unique motivational strength. See Smith
(1995), p. 124.
12 It may be worth noting that this phrase “ ˜perfectly rational™ “ as it occurs in real conver-
sation, does not typically indicate anything unique. Rather, it simply indicates that there
is nothing irrational about a certain action or choice.
13 Worries that this latitude sanctions radically intransitive preferences, transforming agents
into potential ˜value pumps,™ are addressed towards the end of chapter 7. There I explain
and defend the normative signi¬cance of a wide variety of formal restrictions.

118
The relation to other views

psychologically impossible actions cannot be morally required. But if this
particular claim is true, it is plausibly a result of the special link between
violations of moral requirements and liability to punishment, and of the
practical purposes that punishment serves. On the other hand, and as far as
rationality is concerned, there is little reason to think that psychologically
impossible actions cannot be required. For example, a compulsive gambler,
in the middle of a session, may be rationally required to stop gambling,
even though we can predict, as well as we can predict any human behavior,
that he will not stop.
Perhaps we continue to regard certain actions as rational despite their
seeming virtually psychologically impossible because, for practical pur-
poses, it is always at least possible that a given consideration will provide
an agent with her motive for acting.14 But whatever the explanation, it
remains true that we can and do make such claims all the time. No matter
how surprising someone™s action would be, as long as it would fall within
a certain range of possibilities, we do not say that it would be irrational
for the agent to do it. This is one reason why it is a mistake to try to make
the strength of a reason depend upon the strength of the motive it actually
supplies to an agent on a given occasion. This same mistake also arises
when one tries to make reasons relative to something more stable than a
passing desire: for example, to a person™s values or practical identity.15 All
such strategies unreasonably narrow the range of actions of which we can
correctly say that the agent could perform them (or could have performed
them), and be (or have been) acting rationally. Accounts of this kind force
us to call actions irrational if they are contrary to the stable values or
practical identity of an agent. But we need not “ indeed, we would not
and should not “ count the uncharacteristic altruism of a characteristically
egoistic person as irrational.


14 That is to say, even an account of practical reasons that simply stipulates that certain
substantive considerations provide reasons for action will meet what we might call the
˜potential explanation™ requirement on practical reasons. Of course this does not mean
that any given substantive consideration can be reasonably regarded as reason-giving, for
there are many other criteria to be met. For a good recent discussion of the explanatory
requirement, see Johnson (1999), pp. 58“59.
15 See, e.g., Copp (1995), pp. 172“85 and Korsgaard (1996a), chs. 3 and 4. Korsgaard may
perhaps be able escape this problem by means of her claim (p. 102) that one only has a
reason to avoid an action if the performance of that action actually threatens to destroy
one™s identity. But as she herself admits, this escape comes at a heavy price. It means that
one may have no reason of any sort to refrain from even extremely irrational-seeming
actions, as long as one™s character is suf¬ciently resilient.

119
Brute Rationality

h ow m i g h t th e as sum p t i on b e fal se ?
It may seem that the very idea of an agent who is ideal or perfect in
the relevant respects implies that such an agent would be motivated to
some speci¬c degree by any given consideration. But this is not true. In
order to see how the assumption might be false, we will have to address
the naturalistic and nonnaturalistic versions of the ideal motive account
separately.
Consider the following simpli¬ed ideal motive account of normative
reasons
SIM A fact F is a reason for an agent A to in circumstances
C iff F could motivate A to in C, iff A were perfectly
rational.16
SIM is a nonnaturalistic version of the ideal motive account, making
use of an ineliminably normative notion of perfect rationality. It may seem
impossible that a perfectly rational agent would not be motivated to some
speci¬c degree (the ˜most rational™ degree) by any given consideration bear-
ing on his choice of action, in any given set of circumstances. And thus, on
an ideal motive view, it may seem necessary that the strength of a reason
can be given by a single value, perhaps with the admission of some degree
of vagueness. But both of these claims are mistaken. One source of these
mistakes is the view that perfect rationality is a special sort of psychological
state that determines all behavior. But perfect rationality is not a psycho-
logical state, especially on nonnaturalistic ideal motive accounts. Rather,
it is a normative status, like being a perfect driver. Must we assume that
a perfect driver would drive at some speci¬c speed on any given stretch of
road? Must we assume this, even given speci¬c weather and traf¬c condi-
tions, etc.? No. Being a perfect driver is a matter of not breaking certain
rules, and does not fully determine one™s driving style. One of these rules
speci¬es a range of acceptable speeds on any given stretch of road. A per-
fect driver will not exceed the upper limit, or go slower than the lower
limit.17 In the same way, perfect rationality may also be largely a matter
16 Because the point of the current chapter is not to advance any particular version of SIM, no
detailed account of ˜perfect rationality™ is offered here. Rather, the discussion will rely on
particular intuitively plausible claims about the rational status of various actions. It should
be fairly clear that the particular judgments that are used in the following discussion are
ones that would be assented to by virtually anyone who did not already have a sophisticated
philosophical view of practical rationality.
17 Of course, this is a little simplistic, since there are times when a good driver would break
the speed limit, and in general the law permits such exceptions.

120
The relation to other views

of not violating certain principles, such as principle P of chapter 3. If so,
perfect rationality would not ¬x the degree of motivation that a given
consideration would produce.18
Thus we can understand how the uniqueness assumption might be false
on a nonnaturalistic version of the ideal motive account. Consider now a
naturalistic version of the ideal motive account. Such an account claims
that a consideration is a reason for an agent to perform an action in certain
circumstances if that consideration could cause an agent to be motivated

<<

. 20
( 38 .)



>>