Rational Requirements and Practical Reasoning

advertisement

Instrumental Rationality

[DRAFT — Please do not quote]

One ought to take the means to one’s ends. For example, if Fred intends to make an omelette but has no inclination at all to break any eggs, then something seems to be wrong with him: he is practically irrational. Such verdicts of practical irrationality appear to have a distinctive source. In particular, they appear to be independent of whether we are justified in having the corresponding means in the first place: even if Fred has no good reason to make an omelette, it seems that he is still being irrational if he intends to make an omelette and yet lacks any inclination to break any eggs. In this paper I argue that this appearance is correct.

The normativity of instrumental rationality does indeed have a distinctive source. The normativity of instrumental rationality derives, rather, from certain constitutive facts about action — specifically, from facts about the relation between a complex action and its parts or stages.

1. Introduction

Our thought and conduct are, insofar as we are rational and sufficiently well-informed, shaped by instrumental facts — by facts about what the means to our ends are. In this paper I want to examine how exactly instrumental facts bear on our practical thought.

Suppose that one of Fred’s ends is making an omelette. Moreover, suppose that

Fred knows that you cannot make an omelette without breaking some eggs. How does this bear on Fred’s actions? It is certainly not necessary that Fred will break some eggs: Fred might be too engrossed in his favourite TV show to do so. Nevertheless, if Fred does not break any eggs, it is natural to say that, other things being equal, he is being practically irrational.

Fred’s conduct leaves him open to a distinctive sort of blame, precisely in that he fails to take the means necessary to his end.

To see that the blame that attaches to instrumental failures is indeed distinctive, consider a different case. Suppose that Mary pursues an end that she has conclusive reason not to pursue. For example, suppose that one of Mary’s ends is to smoke a cigarette.

Suppose, moreover, that she also believes that in order to smoke a cigarette she needs to buy a pack of cigarettes. Given all this, if Mary does not buy a pack of cigarettes even though

1

the opportunity presents itself to her, she is open to blame in just the same way as Fred.

The important point here is that Mary’s blameworthiness is not due to her acting in a way that is inconsistent with what she has most reason to do: by hypothesis, Mary has conclusive reason not to smoke, and so — presumably — the balance of reasons does not speak in favour of her buying a pack of cigarettes. Examples like this suggest that instrumental failures carry a sort of blame which is independent of the reasons one might have to pursue the corresponding ends.

In this paper I am going to argue that this is indeed so: the requirement of instrumental rationality does indeed have a distinctive source. This view contrasts with the nowadays popular view that what goes wrong in such cases is that the agent is guilty of some kind of incoherence. As I will argue, this view is mistaken: the requirement of instrumental rationality is not just a requirement of coherence. It derives, rather, from some basic — even humdrum — facts about the relation of a non-basic action and its parts or stages. In a nutshell, my claim will be that what goes wrong with Fred if he does not break any eggs is merely that he is doing a really bad job of omelette-making. Likewise, what goes wrong with

Mary if she does not buy a pack of cigarettes is simply that she does a really bad job in her project of smoking a cigarette. We need not look beyond these entirely obvious facts to discover the source of the normativity of the instrumental requirement. (In a way, then, my argument will be that the instrumental requirement turns out to be rather boring.)

Now, the requirement of instrumental rationality is one example of what are sometimes called “rational requirements”. [SLIDE] Rational requirements in this technical sense are constraints on our thought and conduct which hold even in abstraction from the worldly facts that constitute the evidence for our beliefs and the reasons for our actions. For example, logical rules of inference such as modus ponens seem to impose such constraints on

2

our beliefs: it is a rational failure of some sort to believe both p and if p then q but not believe

q, regardless of whether one’s evidence for p and if p then q is any good. All such cases raise questions analogous to the one I aim to address here: how — if not in virtue of the worldly facts that constitute the evidence for our beliefs and the reasons for our actions — do such rational requirements get their normative grip on our thought and conduct?

I am inclined to think that the idea I plan to explore here — namely, the idea that instrumental rationality derives from constitutive facts about action — has analogues in all such cases. An account of the nature of belief, for example, should entail that belief-systems that violate logical constraints are ipso facto defective. I will not be able to argue for this general point in this paper. Even if it is false, what I have to say about the instrumental requirement in particular should not be undermined.

2.

Wide or Narrow Scope?

I have so far rested the claim that the normativity of the instrumental requirement is distinctive on intuitive judgments about cases. It is natural to think that all is not well with

Mary if she intends to smoke but has no intention of buying cigarettes. Since Mary has conclusive reason not to smoke, I suggested that the blame that attaches to her cannot be due to her failing to do something that she has reason to do: it must, it seems, have a different source. However, alternative diagnoses of such cases also seem possible. In particular, it has been suggested that the blame that attaches to Mary is due to her failing to do something she has reason to do: namely, maintain a kind of coherence among her attitudes and actions. (This view is found in earlier work by John Broome, who, however, no longer supports it. See (Broome 1999), (Broome 2001).) Before going into my positive proposal, I would like to explain why I think this view is incorrect.

3

According to this view, the instrumental requirement is a “wide scope” requirement.

The thought here is that instrumental rationality does not require of Mary to perform any particular action; rather (holding fixed Mary’s background belief that she must buy a pack of cigarettes if she is to smoke) instrumental rationality merely requires of Mary to satisfy a disjunction: either to buy a pack of cigarettes, or not to have the end of smoking a cigarette.

That requirement, in turn, is supposed to be one that we have conclusive reason to conform to.

More generally, rational requirements on this view are always requirements of

coherence, broadly understood. On their own, they never require of one to do any particular thing; they only require of one to bring one’s attitudes and actions into a certain kind of harmony, in any one of a number of different ways. In this respect all rational requirements are like the requirements imposed on our beliefs by logical rules of inference. Supposing that I believe both that p and that if p then q, it seems plausible that I can satisfy the rule of

modus ponens either by believing that q or by giving up either one of the beliefs I started out with. Similarly, it is claimed, supposing that my end is to do A, I can satisfy the instrumental requirement either by taking the necessary means to doing A or by abandoning the end of doing A.

There has been a lot of discussion of this proposal in the literature. I cannot survey this debate now; I will simply indicate what I find problematic about the wide scope view.

The problem with the wide scope view, I believe, is simply that it is not the case that rational requirements are, in general, simply requirements of coherence. Intuitively, many rational requirements, including the instrumental requirement, require us to do particular things

— and not simply to maintain coherence. This is perhaps easiest to see with the requirements we are supposed to be under in virtue of our normative beliefs — beliefs about

4

what we ought to do. Switching to a new example, suppose that Mary believes that she ought to go for a run. On the wide scope view, in virtue of this belief Mary is under a requirement which she can satisfy either by going for a run or by dropping her belief that she ought to go for a run. Surely, however, something would be wrong with Mary if she responded in such a situation by dropping her belief that she ought to go for a run. It seems clearly false that she can respond correctly to the requirement she is under in virtue of believing that she ought to go for a run by dropping that very belief. Dropping that belief might be a way for her to escape the requirement; it is certainly not a way for her to satisfy it.

1

Thus there are at least some rational requirements which are not happily read as having wide scope. This is obviously consistent with allowing that some other rational requirements are happily read as having wide scope. Still, the fact that not all rational requirements take wide scope suffices to show that the normativity of rational requirements does not always derive from a standing reason to seek coherence. (I believe that (Kolodny

2005) is exactly right about this.)

1 (Brunero 2010) argues against the narrow scope reading of the requirements we are under in virtue of our normative beliefs. His argument rests on cases with the following structure. Suppose that

Mary believes that she ought to do A, and also that she ought not to believe to do A. According to a narrow scope reading of the corresponding rational requirements, it is still the case that rationality requires of Mary to intend to do A. But this, Brunero suggests, is implausible. However, I believe that the claim that rationality cannot require of Mary to do A in such a case is not incontrovertible.

Whether the claim is true or not would depend on what exactly it is for rationality to require of Mary to do something. Brunero clearly believes that on no reasonable interpretation of rational is Mary required to do A under the circumstances, but this is hardly conclusive. On a sufficiently unambitious construal of rational requirements — such as the one I am developing here — I see no reason why this should be true. (ALSO, I WONDER WHETHER BRUNERO’S ARGUMENT

REALLY SUPPORTS A WIDE-SCOPE READING. DOES MORAL PARTICULARISM — HIS

MODEL — REALLY REQUIRE A WIDE-SCOPE CONSTRUAL OF MORAL

REQUIREMENTS?)

5

To return to our primary topic, I think that the instrumental requirement itself is not happily read as having wide scope. Suppose that Mary’s end is to smoke a cigarette. The claim that Mary can respond correctly to the requirement she is under in virtue of having the end of smoking a cigarette by dropping that very end seems intuitively false. Of course, since — as we have assumed — Mary does have conclusive reason not to smoke, there is a sense in which it would be rational of her to give up her end of smoking a cigarette. But the sense in which this would be rational is surely not that it would be a correct response to the

instrumental requirement.

We can turn this into an explicit argument, from a premise Broome seems to accept.

In a general discussion of reasoning, he asserts that a piece of reasoning is correct only if it achieves the result of satisfying a relevant rational requirement (Broome 2001, 180). This seems natural: the role of rational requirements is to shape reasoning. Contraposing

Broome’s conditional, we have the following [SLIDE]:

(1) If one’s performance is — intuitively — not a correct piece of reasoning, it cannot achieve the result of bringing one to satisfy a relevant rational requirement.

Since Mary’s giving up her end of smoking a cigarette in the situation as described above is not a correct piece of reasoning, it follows that it cannot be a way for her to satisfy any relevant rational requirement. The requirement Mary is under in virtue of having the end of smoking a cigarette does not have wide scope.

Perhaps one might object to this argument, on the grounds that Mary’s giving up of her end to smoke a cigarette is not a correct piece of reasoning simply because it is not a piece of reasoning at all. After all, on Broome’s account, reasoning is supposed to take one from contentful states to contentful states, and it is unclear which contentful state Mary’s

6

giving up of her end to smoke a cigarette is supposed to take off from. In effect, a defender of the wide scope view could argue that the negation in the antecedent of the contrapositive of Broome’s conditional should be internal, not external [SLIDE]:

(2) If a piece of reasoning is — intuitively — incorrect, it cannot achieve the result of bringing one to satisfy a relevant rational requirement.

However, we can overcome this defence simply by switching to yet another example.

Suppose Mary has the end of becoming a more confident public speaker, and she believes that in order to do this she must get rid of her belief that her ears stick out in a funny way.

In this case, at any given time t Mary has failed to take the means to her end if and only if she believes that her ears stick out in a funny way. So the problem identified earlier does not arise: Mary could for all that the principle that reasoning takes off from contentful states has to say on the matter, at least — reason from her belief that her ears stick out in a funny way to abandoning her end of becoming a more confident public speaker. This, however, would not be a correct piece of reasoning. So, by the contrapositive of Broome’s conditional, it would not be a way for her to satisfy a rational requirement.

2

2 My treatment of the case depends on a formulation of the instrumental requirement that differs from Broome’s own. According to Broome, the requirement requires that if you intend E, and believe that M is a necessary means to E, you intend M. On this formulation cases like the above cannot arise, because instrumental failures always consist simply in the lack of an intention. But the lack of an intention is not a contentful state, and so not a fit starting point for reasoning.

I do not find Broome’s formulation of the instrumental principle intuitively very appealing; it seems doubtful to me that there might be constraints on our intentions which are not directly derivative of constraints on our actions. Even if Broome’s formulation is accepted, however, we can still find other examples of rational requirements that would raise the same problem. For example, consider the requirement that if you believe you lack evidence for p, then you do not believe that p.

Intuitively, you cannot reason from your belief in p to the conclusion that you do not lack evidence for p. (see also (Kolodny 2005, 528-530), for a similar argument.)

7

The conclusion we should draw is that the instrumental requirement is not a wide scope requirement. It does not require us merely to achieve or maintain coherence in our practical thought and conduct; it requires us to take the means to our ends. If this is correct, then even if we do have a standing reason to seek coherence, that is not the source of the normativity of the instrumental requirement.

3. Instrumental Rationality and Action

I have so far avoided being pinned down on any very precise statement of the instrumental requirement, but for the purposes of giving a positive account of its source it will be helpful to have such a statement in view. I will adopt the following:

(3) If one is doing a non-basic action A and M is a necessary means to A, then one ought to do M.

Some comments about this formulation are in order. A and M in this formulation range over actions, or things that can be done. In the sense relevant to our concerns, a “basic” action is an action that one performs without taking any means, and accordingly a “nonbasic” action is an action that one performs by performing other actions as means. The notion of means I rely on here is vague but intuitive. For M to be a means to A it is not enough that doing M be somehow involved in doing A; the relation of M to A must also be

represented by the subject, and that representation must be involved in her doing M. Thus, a person with normal motor control over her body can lift her arm as a basic action. Even though the action of lifting one’s arm would seem to have parts which are also actions (for example, contracting one’s relevant muscles), one normally does not need to think of each of any these individual parts and their relation to the whole action. Thus they are not means to lifting one’s arm. By contrast, breaking some eggs is a means to making an omelette.

8

In the sense relevant here, to say that M is a necessary means to A does not mean that it is physically impossible for one to do A without doing M. Rather, it means that, taking into account only courses of action that are realistically open to one — given one’s priorities, as well as one’s restrictions of time and resources — all those courses of action that result in one’s doing A include one’s doing M. This is the sense in which it is necessary for Mary to buy a pack of cigarettes if she is to smoke.

One way in which my formulation of the instrumental requirement might appear unusual is that it is formulated in terms of doing A, rather than simply having A as an end, or intending to do A. To some extent, this appearance is illusory. Even though my formulation does not explicitly refer to an intention to do A, a closely related intention is clearly lurking in the background: if one is doing A intentionally, one must have an intention to be doing A. In current terminology, this is a progressive intention, or an intention-in-action. Of course it is overwhelmingly plausible that a version of the instrumental requirement applies to prospective intentions, or intentions for the future as well. Such cases, however, introduce complications which would only distract us here. In any case, as John McDowell, among others, has urged, intentions in action might in fact be the philosophically prior topic, with intentions for the future to be understood as the same sort of thing, only before the time for their execution has arrived. Hopefully, then, this restriction of the scope of my discussion won’t be too damaging.

3

3 The prospective case introduces some complications. For one thing, we must allow for cases in which the agent has only the vaguest idea of what will constitute the means to her end: perhaps a course of action prospectively intended as the complete means to doing A can include actions like

“improvising along the way”. Moreover, we must also allow for mistakes: an agent will not always be right about what constitutes the means to her ends. I will have to ignore such complications here.

9

Finally, many authors qualify the instrumental requirement, adding in the antecedent the condition that the agent believe that M is a necessary means to A. I have omitted this extra condition, mainly for simplicity, but also because I do not find the reasons for it compelling. Including it would not have made a significant difference in what follows.

For our purposes, the crucial point about this formulation of the instrumental requirement is that it has a narrow scope “ought” in its consequent, and therefore it raises the questions that constitute the central topic of this paper.

My central claim in this paper is that this instrumental requirement can be derived from constitutive facts about action. What’s more, this derivation sheds light on the distinctive kind of “ought” involved in the instrumental requirement.

The fundamental point to notice here is the following [SLIDE]:

(4) For any action A, if one is doing A then one is engaged in a process which, if all goes well, will result in one’s having done A.

If A is a non-basic action, that process will itself be a sequence of further actions one performs, and which together constitute sufficient means to doing A. Of course to say that a sequence of actions constitutes sufficient means to doing A does not entail that performing all actions in that sequence guarantees success in doing A: for example, picking up the gun, taking aim and shooting is a sufficient means to hitting a bulls-eye, but it does not guarantee that one will hit a bulls-eye.

Now, although this principle is very intuitive, there are putative counterexamples.

For example, suppose that, after having prepared all of the ingredients, Fred is simply hanging around waiting for the oil in the pan to reach the right temperature. Although it is

10

clearly true of Fred right now that he is engaged in the non-basic action of making an omelette, one might be tempted to say that it is not true of him that he is engaged in any further action that is a means to omelette making.

I believe such putative counterexamples can be resisted. For one thing, in many cases we will be able to find actions which one is performing, and which can be described as means to performing the relevant non-basic action. For example, while Fred is apparently just hanging around, it is still true of him that he is heating up the oil — and this is clearly a means to making an omelette. I expect that many putative counterexamples will be amenable to this sort of response — some times, perhaps, at the cost of some artificiality.

Cases for which this response fails, I suggest, will be cases of “interrupted” actions.

Suppose that Mary is writing an article on nudibranchs. She spends some time thinking about nudibranchs, some time researching nudibranchs in their natural environment, some time experimenting in her laboratory, some time typing in front of her computer. But she also spends a lot of time doing other things: going to the movies, teaching her classes, or sleeping. So suppose that right now Mary is asleep. Although there does seem to be a sense in which we can still say of Mary that she is writing an article on nudibranchs, there is also a clear sense in which we can say that she is not, right now, writing anything. Writing an article is the sort of action that admits of interruptions. If A admits of interruptions, then during one of those interruptions one might still be doing A in one sense, while not be doing A in another sense — and it will be this second, intuitively more robust sense of “doing”, that is captured by our principle.

Thus, in this robust sense of “doing”, unless one is engaged in a process which is a sufficient means to doing a non-basic action A, one is not doing A. This fundamental

11

principle about non-basic action, I think, is the root from which the instrumental requirement ultimately derives.

The crucial piece of work that still remains to be done involves the normativity of the instrumental requirement. As I have already suggested, I believe the instrumental requirement ultimately derives from the facts about action that (4) aims to capture. Notice, however, (4) is not explicitly normative. Still, I believe that we have the materials to derive a genuinely normative requirement from it. (This part of my paper draws heavily on the work of Michael Thompson.)

To begin with, let us consider quite generally the relation between one’s doing A and one’s having done A — the relation between a sentence in the progressive and what we may call (following Zoltan Szabo (2004)) its perfective correlate. Consider Mary, a young child engaged in the process of growing up. This fact about Mary obviously does not entail that it is necessary that at any point in the future Mary will have grown up: all sorts of things might happen to prevent this state of affairs from obtaining. Nevertheless, it seems clear that there is a certain sense of normality such that the fact that Mary is now engaged in the process of development serves to pick out certain future courses of events as in that sense normal. Given a normal run of events, Mary will have grown up in a few years time. Notice that — as

Thompson stresses in a similar context — this sense of normality is not merely statistical normality. Even if a deadly global pandemic has made it highly unlikely that either Mary or any of her peers will ever grow up, there is a clear sense in which the normal course of events still involves the state of affairs of Mary’s having grown up.

Since the sense of normality invoked here is not statistical normality, my claim allows that one might be engaged in processes that one has only a very slim chance of completing.

12

Nevertheless, in its unqualified form my claim about processes seems to entail that one cannot be engaged in the process of doing A unless it is possible for one to have done A. It is not clear that this is an acceptable consequence. Szabó (2004) uses this point in his elaborate critique of attempts to reduce the progressive to the perfective. His example — which seems intuitively correct to me — involves a medieval architect who is, right now, building a cathedral, although there is no possible future course of events which involves his having built the cathedral (the building of the cathedral, we may suppose, takes two or three hundred years, and the labour of many successive architects, to complete). Since my project here is not the reduction of the progressive, I — unlike the semanticists Szabo criticizes — am within my rights to simply qualify my claim. So my claim is the following [SLIDE]:

(5) If one is doing A, then — assuming that it is possible for one to complete

A — there is a sense of normality such that, in a normal run of events, at some future time one has done A.

A full treatment of the metaphysics of processes would have to do away with this inelegant qualification, but I cannot undertake this task here.

Now, even though the sort of normality involved here is not merely statistical normality, it does not follow that it involves any kind of normativity. In the example above the normality involved was in some sense biological normality. It is unclear whether we should speak of normativity in such a case: although the notion of biological norms has currency, it is not at all clear that these norms are really normative. Moreover, these remarks about processes are meant to be entirely general. It is a general fact about processes that they induce a sorting of future courses of events into the “normal” ones and the rest. This holds of processes as completely free of the normative as a stone’s falling to the ground.

The fact that the stone is falling to the ground induces a sorting of future courses of events

13

into the normal ones, in which the stone reaches the ground, and the rest — in which something interferes and the stone never reaches the ground. Obviously, this sorting is not normative: certainly the stone is not to be blamed in those courses of events in which it fails to hit the ground.

Suppose now that we consider a special class of processes — namely, those that are intentional actions. Consider, once again, Fred’s omelette-making. If Fred is making an omelette, then — according to what we just said — this fact induces a sorting of future course of events into the normal ones, which include Fred’s having made an omelette, and the abnormal ones, in which Fred for one reason or other fails to make an omelette. In all these cases we may speak of Fred’s practical failure.

It is clear that not all practical failures carry any sort of blame — so the concept of a practical failure as such is not a normative concept. For example, Fred might fail to complete his omelette because his apartment is blown up by terrorists. Nevertheless, it is also clear that some practical failures do carry a certain sort of blame. If Fred’s omelettemaking fails due to a factor Fred has control over, then often the practical failure can be

blamed on him. Such practical failures are, in this respect, radically different from the failure of the stone to hit the ground in an earlier example, or from Mary’s failure to reach adulthood, or even from Fred’s failure to make an omelette due to the terrorists’ blowing up his apartment. For example, if Fred is just too engrossed in his favourite TV show to get up out of the couch and get down to breaking some eggs, it seems that his failure at omelettemaking is due to him — it was something about the state of his will at the time that made it the case that the omelette did not get made. In such a case, it is natural to accuse Fred of some sort of practical irrationality. On my understanding, the point of this type of accusation of practical irrationality is precisely to pick out a kind of practical failure that is

14

due to a defect in Fred’s state at the time, rather than bad luck or external interference.

Clearly, one might be guilty of this kind of practical irrationality even if the practical failure affects a project that one had no good reason to pursue in the first place.

With this account of practical irrationality in place, let us finally return to our primary topic, namely the requirement of instrumental rationality itself. Suppose that one is doing A, and

M is a necessary means to A. Suppose, moreover, that one recognizes that M is a necessary means to A, and that doing M is within one’s powers. Now, the fact that one is doing A induces, according to what I argued above, a sorting of future courses of events into the normal ones — which include one’s having done A — and the rest. Moreover, since M is a necessary means to A, no course of events which does not include one’s doing M is a normal course of events — in no course of events in which one fails to do M does one succeed in doing A. Cases in which one fails to do M are all cases of practical failure. Finally, given our assumption that one recognizes that M is a necessary means to A, and that M is within one’s powers, it seems plausible that, in at least some of these cases, one is to be blamed for this practical failure — in particular, in some of these cases one is open to a charge of practical

irrationality. In those cases, one’s problem is precisely that one has fallen afoul of the requirement of instrumental rationality.

References

Broome, John. 1999. Normative Requirements. Ratio XII, no. 4: 398-418.

———. 2001. Normative Practical Reasoning. Aristotelian Society Supplementary Volumes 75:

175-193.

Brunero, John. 2010. The Scope of Rational Requirements. The Philosophical Quarterly 60, no.

15

238: 28-48.

Kolodny, Nico. 2005. Why Be Rational? Mind 114: 509-563.

Szabó, Zoltán. 2004. On the Progressive and the Perfective. Noûs 38: 29-59.

Thompson, Michael. 2007. Lige and Action. Cambridge, MA: Harvard University Press.

16

Download