reasoning, practical
Figuring out what to do; reasoning directed towards action (as contrasted with reasoning directed towards arriving at belief).

Practical reasoning is a rapidly changing area of study; this article describes the state of the field as it has shaped up over the 1980s and 1990s. For earlier work, see Raz (1978).
The current debate in practical reasoning focuses on the question: what inference patterns are legitimate methods of arriving at decisions or intentions to act, or other characteristically practical predecessors of actions such as evaluations, plans, policies, and judgments about what one ought to do? The spectrum of competing theories ranges from the very minimal, allowing only one form of practical inference (or even none), to maximally permissive views that "let a thousand flowers bloom." The remainder of this article surveys the most prominent positions on this spectrum. Other important questions in the field include: What mental states and processes are involved in practical reasoning? Is there a principled distinction to be made between practical and theoretical reasoning (that is, reasoning directed towards belief rather than action)? How can one argue about what putative forms of practical inference are legitimate, i.e., are really inference? Is there a fixed list of forms of practical inference to be discovered, or can methods of practical reasoning be invented?
Nihilism about practical reasoning is the view that there are no legitimate forms of practical inference, and that consequently there is no such thing as practical reasoning: appearances notwithstanding, there is no mental activity that counts as figuring out what to do. This position is the most minimal on the spectrum of views about practical reasoning; it suffers from a shortage of contemporary defenders, but was argued for by Hume (1739/1978, pp. 413-418, 456-470). (Nihilism is canvassed as a possible form of irrationality by Korsgaard (1986, sec. 3). For discussion of Hume, see Hampton (1995) and Millgram (1995).) Nihilism was probably entailed by early twentieth-century noncognitivism in metaethics, in particular by emotivism; however, it was not discussed at the time. Nihilism should be regarded as the null hypothesis against which other accounts of practical reasoning must be defended.
Instrumentalism holds that all practical reasoning is means-end reasoning, that is, that figuring out what to do is entirely a matter of determining how to achieve one's goals or satisfy one's desires. Such reasoning may consist in finding causes for the outcomes one wants; but it may also involve -- for instance -- scheduling, or picking one among the various options that would achieve one's goals. (Instrumentalism in this broad sense -- as opposed to the sense in which it is restricted to finding causes for outcomes -- has sometimes also been called "internalism about reasons for action;" in the last few years, however, usage has been shifting to make internalism a distinct notion.) Because infinite regresses are generally thought not possible in instrumental justification (finite creatures have only finitely many suitably distinct desires), and because circularity in instrumental justification is thought to be unreasonable, the instrumentalist position usually has it that practical justification bottoms out in desires one just has: you can reason about how to get what you want, but not about what to want in the first place.
Instrumentalism is the default view in the field, and probably among philosophers in general. However, despite its pervasiveness, there is very little explicit argument for it. Hume (1777/1975, p. 293) is perhaps the locus classicus of the view, which is -- despite Hume's apparent commitment elsewhere to nihilism about practical reasoning -- often called Humeanism. Smith (1987) attempts to tie instrumentalism to belief-desire psychology (see folk psychology) and to an understanding of beliefs and desires in terms of direction of fit. Williams (1981) argues that reasons for action must invoke desires on the grounds that only desires, broadly construed, can explain actions, and that reasons for action must be able, in suitable circumstances, to explain actions.
The shortfall of argument notwithstanding, there are a number of things to be said for the position. First, means-end reasoning is perhaps the least controversial form that practical reasoning has been alleged to take: we all know what it is like to try to figure out how to achieve a goal, and it is often straightforward to determine whether a mistake has been made. Second, since instrumental reasoning proceeds from desires you have, your stake in the results of such reasoning seems very clear -- the actions it tells you to take promise to get you things you already want. Third, when people do seem to act for reasons, it is generally possible to ascribe to them an appropriate desire. Fourth, non-instrumental practical reasoning would evidently allow one's ultimate desires to be corrected by others; but we are familiar with such attempts at correction, and they as a rule are paternalistic, heavy-handed, dogmatic, and unconvincing. Accordingly, instrumentalists insist that the burden of proof lies with their opponents.
Whether or not instrumental reasoning is the only kind of practical reasoning there is, there is widespread agreement that it is a kind of practical reasoning. (Thompson (1998) is a very interesting dissenting voice.) But although it is uncontroversial in this respect, it is by no means well-understood. The central problem is the defeasibility of instrumental reasoning, that is, the fact that an apparently satisfactory instrumental inference can be defeated by adding further premises, and that we have no means of specifying the defeating conditions up front. For instance, if my end is to have an espresso, a suitable means might be going to a particular cafe... but I might quite properly retract my decision when I learn that its management donates a percentage of the cafe's profits to a terrorist group. There are obviously indefinitely many defeating conditions of this kind, and so we do not know how to say under what circumstances a means-end inference ought to go through. Whether or not instrumentalism is correct, perhaps the most important advance that could be made in the field would be to figure out how means-end reasoning really works.
In 1976, Jimmy Carter ran for President on the slogan: "Why not the best?" One answer, given some time back by Herbert Simon (1957, chs. 14, 15; see also 1979), is that finding the best choice can be computationally and otherwise too expensive. Another is that choice sets may fail to contain maximal elements; when there is no best, choosing the best is not an option (Landesman 1995, Fehige 1994). The alternative to maximizing is satisficing, that is, choosing an option that is, while perhaps not the best, good enough. Satisficing is naturally thought of as instrumentalist in spirit, and defended as a kind of second-level maximizing: the best first-level strategy, once typical information costs are taken into account, may be satisficing rather than maximizing. However, satisficing has also been defended as simply being in line with our intuitions about the rationality of particular choices (Slote 1989).
Reasoning with maieutic ends
A standard objection to instrumentalism is that it makes ultimate ends come out arbitrary: your ultimate ends are the things you just happen to want, they are beyond the reach of deliberation and rational control, and we know from experience that this is unrealistic. A response to the objection can stay within the spirit of instrumentalism by appealing to maieutic ends (also called second-order ends or second-order desires), that is, ends that consist in having other ends or desires. For example, you might want to have a career in medicine for entirely financial reasons; in order to have the career, you have to care about the right things, e.g., healing the sick; so you come to want to have the end of healing the sick. As the example suggests, it is possible to reason about the desirability of wanting something, without expanding the repertoire of inference patterns beyond the instrumental. One can (it is held) adopt desires that are ultimate in the sense that their objects are not wanted as means to further ends; but one can adopt these desires for instrumental reasons, because although the desires' objects are not wanted as means to further ends, the desires themselves are so wanted. This approach is developed and defended by Schmidtz (1995).
Plans and intentions
On the planning view of rational deliberation advanced by Bratman (1987; see Bratman 1983 for an early but more concise presentation of the view), practical reasoning consists largely in the adoption, filling in, and reconsideration of intentions and plans. (In Bratman's usage, a plan is a more elaborate and developed intention, and an intention is a small plan.) Plans have two important characteristics that distinguish them from desires. First, they are typically incomplete: your plan for flying to Spain may include the intention of getting to the airport, but until the day arrives it may well not include a subplan for getting to the airport e.g., calling a taxi, waiting for it, getting in, and taking it there. Second, plans are stable: normally, one reasons about how to execute and fill in one's plan, but not, unless special circumstances arise, about whether to reject the plan in favor of some other.
Practical reasoning that avails itself of plans has advantages over reasoning that uses only beliefs and desires. Because plans are stable, plans make the practical reasoning one has to do manageable by framing one's deliberations, and so restricting the number of options that need to be considered. For instance, in considering how to get to the airport, you must weigh the merits of driving, taking a cab, taking the bus, and getting a friend to drop you off; but because the plan to go to Spain frames your deliberations, you need not consider the option of going to Peru instead. Because plans are usually filled in as needed, they can efficiently take account of information that becomes available later rather than earlier: when you formed the intention of going to Spain, you could not know whether your spouse would want to use the car that day. Because plans are stable, they can facilitate interpersonal coordination: knowing that Joe plans to be at the cafe at 3:00 is an entirely different matter from knowing that he wants to be there at 3:00; I am much more willing to go there to meet him on the former basis than on the latter, because I know he will not reconsider unless special circumstances arise. Because plans are stable, they also facilitate intrapersonal coordination, that is, coordination over different times; knowing that you will finish the paper you have now started to write, because your intention to do so is stable, provides some assurance that the effort spent in starting it will not have been wasted.
Bratman's views regarding when reconsideration of plans is rational have been developing over the last decade. One argument turns on the consequences of having the policy that triggers reconsideration: if you must reconsider every time a new bit of information turns up, your decision making task will swamp your cognitive resources. Consequently, the planning theory of practical rationality can deliver prescriptions for action that differ from those of the traditional instrumentalist theory. For example, suppose you adopt a plan because you believe, rationally, that it will best satisfy your desires; suppose circumstances change so that it no longer does, but your rationally held policy for reconsidering plans does not take the change in circumstances to warrant reconsideration. Then the instrumentalist theory may hold that it is rational for you not to perform the actions dictated by the plan, whereas the planning theory may hold that it is rational -- even though executing the plan will not satisfy your desires anymore.
Many of our ends -- runs one objection to instrumentalism -- are simply too vague or indefinite to serve as starting points for means-end reasoning, so practical reasoning must consist partly in further specifying the overly indefinite ends. For instance, I want to improve my looks; but, before actually making purchases at the makeup counter, I need to figure out just what improvement in my looks would be. (The early pivotal papers are by Kolnai (1978) and Wiggins (1991); Kolnai, who thinks that Aristotle was an instrumentalist, develops the view as an alternative to Aristotle's, while Wiggins attributes the specificationist view to Aristotle.) Richardson (1994) advances a further reason for specification of ends. Many of our ends conflict, but often those conflicts (whether between one's own ends, or the ends of different people) can be removed by further specification of the ends in question. Since the point of cospecification is to remove conflict between ends, specifications should be chosen that make the ends cohere with one another (and with other background elements of one's evaluative system).
In contrast to the three immediately preceding positions, which remain instrumentalist in spirit, specificationism is (like the further positions we will survey) a full-fledged alternative to the view that all practical reasoning is means-end reasoning: only when supplemented with the rational specification of ends is instrumental reasoning viable at all. The most important item on the specificationist agenda is to make out what distinguishes correct or rational from incorrect or irrational specifications of an insufficiently definite goal.
Practical contradiction resolution
One important aspect of theoretical reasoning (that is, reasoning directed towards belief) is resolving contradictions in one's system of beliefs. The way in which this is done is not well-understood, but it is nonetheless possible to ask whether it has a practical analog: a form of practical reasoning directed towards resolving (something that would count as) a practical contradiction. It has been suggested by Candace Vogler that such contradictions may be generated by practical versions of so-called "evening star-morning star" cases, as when one wants to visit Siam, and to avoid visiting Thailand, but then becomes aware that Siam is Thailand. Korsgaard (1990) has a useful discussion of the notion of a practical contradiction in Kant.
Coherence-driven reasoning
Practical reasoning is sometimes thought to be a matter of adjusting one's practical take on things, together with one's actions, in the direction of greater coherence. Just what this suggestion comes to will depend both on what the elements of one's practical take on things are thought to be, and on what the coherence of such elements with one another is supposed to consist in. For instance, if preferences are the relevant items, coherence might be understood to consist in the agent's preferences satisfying the conditions for his having a well-defined utility function. (For the canonical account of utility functions, see Luce and Raiffa, 1957, ch. 2.) However, the expected-utility approach to coherence has the problem that no sense is given to the notion of a set of preferences being more or less coherent; your preferences are either coherent (if they induce -- that is, can be represented by -- a utility function), or they are not. The expected-utility approach to coherence specifies an ideal that is unattainable for human beings, without saying what it would be to move closer to it, or farther away.
The elements of one's practical take on things might alternatively be thought to consist in goals, subgoals and actions. In this case, coherence-driven practical deliberation would amount to choosing the subset of the goals and actions under consideration that best cohere with one another. Practical reasoning of this kind can be described as "inference to the most coherent plan" -- in the event that finding the most coherent plan is impractical, the recommendation is to find as coherent a plan as one can. The most urgent issue in this area is the development of comparative notions of coherence that are precise enough to give clear answers to questions of the form: of these competing plans, which is the most coherent? Without notions of coherence that are usable in this way, appeals to coherence are empty, and the merits of coherence-driven accounts of practical reasoning cannot be assessed. One such comparative notion has been modeled computationally, by using quasi-connectionist networks to represent the competing plans. (Thagard and Millgram, 1995; Millgram and Thagard, 1996) Another is specified as a constraint satisfaction problem. (Thagard and Verbeurgt, 1998) Coherence-driven accounts are yet another alternative to the instrumentalist paradigm; on the goal-oriented notion, for example, goals can be adopted on the ground that they cohere with other goals that one already has, even if achieving them would not be a means to any end that one already has. If coherence-driven accounts of practical reasoning are to make headway, improved definitions of coherence and further development of techniques for modeling them are necessary.
Kantian theories of practical reasoning typically require that reasons be universalizable: roughly, that it be possible for everyone in like circumstances to act likewise on the basis of a similar reason. Nell (1975) argues that the requirement imposes substantial constraints on what actions are permissible, and explains how the requirement can be proceduralized. Universalizability acts as a filter through which proposed actions and the reasons for them are passed, but it can also be used to generate reasons on its own, when not acting on a proposed reason would fail the universalizability test. Contemporary interest in universalizability is primarily due to the role it plays in Kantian moral theory, which is today one of the most prominent positions in ethics; Korsgaard (1990) reconstructs Kant's reasons for insisting on the universalizability requirement.
Identity-based practical reasons
Korsgaard has recently suggested that a theory of practical reasoning should make room for a class of reasons that express one's self-conception or "practical identity," "a description under which you value yourself ...find your life to be worth living and your actions worth undertaking" (1996, p. 101), e.g., being a philosophy professor, a Canadian, a "made man," and so on. The appeal to practical identities goes some of the way towards meeting a challenge posed by Williams on behalf of the instrumentalist position, that of showing how the practical reasons of different persons in what are substantially similar situations can vary, without (as the instrumentalist does) simply referring the difference to their differing desires. (Williams 1995, pp. 186-194)
Practical empiricism
On the views we have seen so far, the source of an agent's reasons for action lies within the agent: in his goals or ends or desires (instrumentalism, satisficing, maieutic ends), in his intentions (the planning view, Kantian universalizability), in the ways all these cohere with one another, or in his practical identity. On the instrumentalist view, for instance, experience can supply the facts needed to determine how to attain one's ends, but the ends themselves are set from inside, by one's desires. Against this, practical empiricism has it that it is both possible and necessary to learn what matters, and what is important, from experience. There is no reason to think that goals, priorities, evaluations and other like pieces of an agent's cognitive equipment will be useful guides to action if the world is not allowed to have its say in what they look like. The desires and intentions with which one comes to a situation may be simply irrelevant (likely when the circumstances are novel), or their objects may prove disappointing when obtained; successful agency requires an ability to correct one's assessments and agenda on the fly.
For example, perhaps I originally took climate control in a car to be more important than mechanical reliability; after many unpleasant experiences with mechanics, I conclude that I was mistaken, and that a reliable car is generally to be preferred to an air conditioned one. This suggests that the basis for correction will be a practical analog of observation, and that, because learning from experience requires the ability to generalize from past observations to future instances, practical empiricism should be committed to a practical version of inductive inference, one that moves from particular to general practical judgments. A version of practical empiricism along these lines is developed by Millgram (1997).
Redescription as practical reasoning
In order to draw the right conclusion about what to do, you normally have to proceed on the basis of an adequate description of your situation. Arriving at the description is usually regarded as theoretical rather than practical reasoning; you are reasoning about the facts, rather than about the values. Murdoch (1970) differs on this point: arriving at the description that is ultimately the basis for action is the important and hard part of practical reasoning (in part because facts cannot be distinguished from values -- or, more interestingly, because the attempt to do so is itself the expression of a particular set of values; on this last point, see Diamond 1996). Although Murdoch's work predates the two-decade period under review, her writing is only now coming to be seen as advancing a view about practical reasoning.
Murdoch's discussion focusses almost entirely on one aspect of the process of redescription, that of overcoming the temptation to see situations in emotionally convenient ways. In her most famous example, a mother conquers her jealousy, and learns to see her daughter-in-law as refreshing, simple, spontaneous, and delightfully youthful, rather than vulgar, rude, undignified, and tiresomely juvenile. Her insight is important but incomplete: even if one agrees that redescribing one's situation is practical reasoning, it will be hard to accept that such redescription is all there really is to figuring out what to do; and it will be as hard to agree that all there really is to successful redescription is getting past the emotionally induced distortions.
Other positions
There are a handful of other positions that deserve mention but are difficult to place on the spectrum.
First, there is the common idea that expected-utility theory (see Luce and Raiffa 1957) is a satisfactory rendition of practical reasoning; this is hard to place because the formalism is advanced both as a kind of instrumentalism (where the agent's goal is to maximize his expected utility), and as a formal notion of coherence (see above). What is more, some of the justifications for the coherentist interpretation are meant to be instrumentalist in form and spirit; see, e.g., McClennan (1990). Hampton (1994) contains a recent critical discussion of the instrumentalist interpretation.
As a matter of fact, the formalism may well be compatible with other positions on the spectrum as well. It is certainly compatible with nihilism, and the wide availability of personal computers has made this easy to see. To the computer-literate, the expected-utility formalism looks like a data compression technique, or perhaps an encryption device, rather than a representation of reasoning, or even of rationality in one's preferences. The formalism gives a way of representing one's preferences -- provided that they satisfy a handful of actually quite demanding conditions -- by assigning real numbers to outcomes, such that selecting the action with the highest expected utility (the sum of the products of those numbers and the probabilities of the respective outcomes, given that the action is performed) will be choice conforming to one's preferences. That is, it performs a function analogous to those of the popular ZIP and Compress programs: just as these store your information in a file formatted to take up less space on your hard disk than an uncompressed file, so the expected utility formalism allows you to encode unmanageably many preferences as much more manageable real numbers. Of course, there may be as many numbers as there were preferences in the first place, and, if one is unlucky, they may not be much more manageable than the preferences had been; in that case, one can think of the encoding as encryption rather than compression. But, thought of in either of these two ways, retrieving your original preferences via an expected-utility calculation would be no more practical reasoning than retrieving your original file by decompressing a "zipped" file (or by decrypting an encrypted file) is theoretical reasoning. And the formalism could be used to encode suitably structured preferences even if there was nothing that counted as practical reasoning at all -- that is, even if nihilism were correct.
Second, Velleman (1989; see 1985 for a shorter and easier-to-read discussion) argues that intentions are self-fulfilling predictions (and so practical reasoning is in fact a variety of theoretical reasoning). The predictions are self-fulfilling because agents desire to understand what they are doing, and acting on the basis of such a self-fulfilling prediction produces the requisite kind of self-understanding.
Third, Brandom (1994, pp. 243-253) claims that desire ascriptions merely express commitments to the material correctness of practical inferences that do not themselves involve desires. For instance, to say that someone desires not to get wet is to make explicit his commitment to inferring "I will take my umbrella" from "It is going to rain;" the desire is not itself a suppressed premise of the inference. This amounts to an argument against instrumentalism, to the effect that the instrumentalist position radically misconceives the point of desire ascriptions. The omnipresence of desires served by one's actions is not, as it has been taken to be, evidence for instrumentalism. Rather, that one's practical inferences can generally be recast in a form that invokes desires is entirely neutral with respect to the question: what patterns of practical inference are legitimate?
The state of the field
Two decades ago, practical reasoning was, in the minds of most professional philosophers, not separable as an area of study from ethics, and it tended to be associated with historical scholarship. Instrumentalism was the clearly dominant position, to which a revived Kantian morality was the most visible alternative. Some strong work was being done on prudential (i.e., future-regarding) reasons as well, since these were regarded as a possible model for a treatment of altruistic reasons (Nagel 1970). But, by and large, the field was stagnant.
It has since come a long way. Practical reasoning is no longer the handmaiden of ethics, and today theories of practical reasoning are not normally advanced merely as components of some favored moral theory. The fortification and defense of a very small number of entrenched positions inherited from the great dead philosophers has given way to a healthy profusion of competing and largely new views. Important ideas and arguments turn up annually or semi-annually -- a rate that marks a philosophical subspecialty as rapidly developing.
Work on practical reasoning has consequences for ethics and philosophy of mind: moral reasoning is practical reasoning concerned with moral subject matter; philosophical ontologies of the mental typically reflect whatever happens to be the current view of rationality. (E.g., the popularity of belief-desire psychology is partly attributable to the recently widespread acceptance of instrumentalism, a view in which desires loom very large.) So we can expect recent developments in practical reasoning to produce ripple effects in those areas also.
Elijah Millgram