Time for Action

  1. 1.  New York University

Prologue: “All Out Judgment”

[T]he event whose occurrence makes ‘I turned on the light’ true cannot be called the object, however intentional, of ‘I wanted to turn on the light’. If I turned on the light, then I must have done it at a precise moment, in a particular way—every detail is fixed. But it makes no sense to demand that my want be directed to an action performed at any one moment or done in some unique manner. Any one of an indefinitely large number of actions would satisfy the want and can be considered equally eligible as its object.[2]

All we can judge at the stage of pure intending is the desirability of actions of a sort, and actions of a sort are generally judged on the basis of the aspect that defines the sort. Such judgements, however, do not always lead to reasonable action, or we would be eating everything sweet we could lay our hands on … .[3]

[A]n all-out judgement makes sense only when there is an action present (or past) that is known by acquaintance. Otherwise … the judgement must be general, that is, it must cover all actions of a certain sort, and among these there are bound to be actions some of which are desirable and some not. Yet an intention cannot single out a particular action in an intelligible sense, since it is directed to the future.[4]

What exactly is the problem that is worrying Davidson in these passages? It has something to do with the temporal order of motive and action. He thinks that the mental states by which action is constitutively caused must precede their behavioral effect, and that this necessary temporal order precludes the possibility that these mental states can have as their intentional object the particular action that they cause. But why is this temporal constraint problematic in Davidson’s eyes?

The constraint is problematic, apparently, because it stands in the way of the evaluative judgment that constitutes an intention to act. Davidson conceives of an intention as an “all-out” judgment about the desirability of an action. An all-out judgment is usually preceded by an all-things-considered judgment, which pronounces an action desirable insofar as the available reasons go. The subsequent all-out judgment pronounces the action to be desirable, full stop. Davidson thinks that the temporal order between intention and action precludes the agent from making this judgment about a concrete particular action, because a singular judgment would require the agent to be acquainted with the action, as he cannot be with an action that has not yet occurred. The agent’s intention must therefore pick out the intended action by description, hence as a kind of action rather than a concrete particular. The problem is that the agent cannot specify the intended kind of action with enough specificity to rule out every undesirable member of the kind (e.g., eating a poisoned sweet). Since the agent will never be able to describe a kind of action with enough specificity to rule out undesirable instances, he seems to be stymied unless he leaps to an all-out conclusion for which he lacks sufficient grounds.

One flaw in Davidson’s reasoning—albeit an inconsequential flaw—is that intending a concrete particular action doesn’t require that it already be available as an object of acquaintance. A particular action can also be picked out in advance by definite description. More importantly, it can be picked out in advance by what I would call “proleptic demonstration”.

I perform a proleptic demonstration when I point to something by causing it. For instance, I can tell a kindergartener, “The letter ‘a’ looks like this,” fixing the reference of “this” with the movement of writing an “a”, which directs attention to its own result on the blackboard. Or the referent of “that” can be a stinging sensation pointed out by the slap that causes it: “Take that!” Because a demonstration can direct attention to one of its effects, it can point to what doesn’t yet exist and cannot yet have been known by acquaintance. Hence, the intention to take this step can pick out a concrete particular step by pointing to it, because it can point to the step precisely by causing it.

Davidson’s point still stands, however, since a singular intention that used proleptic demonstration would still have to be based on general grounds, which would be inadequate to rule out undesirable instances. My practical reasoning, when laid out in Aristotelian fashion, cannot go like this.

I want to take a step.

This is a step.

 I’ll take this.

There is nothing to identify as a step until it has been at least intended, and so I cannot intend to take a concrete particular step on the grounds of its being one. If I am going to intend a particular step demonstratively, I’ll have to reason something like this:

I want to take a step.

So I’ll take this step.

Although I can intend a concrete particular action, then, I still have to intend it as a member of a general kind—in this case, a step—differentiated from other members only by being the one produced by this very intention, a distinction that cannot support a comparative judgment in its favor and that consequently leaves open the possibility of its being a undesirable member of the kind.

Yet this risk would not be eliminated, as Davidson seems to think, by acquaintance with the concrete particular action, and here is a more consequential flaw in Davidson’s reasoning. For even after I have performed an action and become acquainted with it, I cannot canvass all of the respects in which it may be undesirable. If it would be undesirable to alert the burglar in the next room who, unbeknownst to me, is rifling through my desk, then even after I have flipped on the light, I may not know everything relevant to the desirability of having done so. The knowledge I gain by being acquainted with the action still leaves open the possibility that it is disastrous, contrary to my all-out judgment of its desirability.

Why does Davidson think that I need to insulate my all-out judgment against falsification? There is no problem with a judgment’s being defeasible even after it has been detached from specific grounds or from the generic hedge “all things considered”. Even an unqualified or categorical judgment can be revised or retracted after the fact: “Flipping the switch would be desirable, full stop”; “Oops, it wasn’t.” “Categorically, eating a sweet would be desirable”; “Yuck—not that one!”

What now emerges is a crucial difference between a desirability judgment and an intention, a difference that raises doubts as to whether the one can constitute the other, as Davidson believes. An intention cannot be retracted after having been executed: once the agent has decided to do something and then done it, he cannot change his mind. How, then, can his intention consist in a judgment that can so easily be revised or retracted after the fact? The role of intention in rational thought is fundamentally different from that of a desirability judgment, whose role allows for subsequent revision or retraction. How can intention consist in a judgment that plays a fundamentally different role?

I suspect that Davidson’s phrase “all-out judgment” is meant to suggest that the judgment is not just unconditional but final, in the sense that it cannot be retracted or revised. An all-out judgment, in Davidson’s mind, is a judgment to which the agent commits himself irrevocably, once and for all. You might think: There is no such thing as a once-and-for-all judgment—no such creature in the mental menagerie. To which the Davidsonian reply is: “Well, there is such a creature, namely, the desirability judgment that constitutes an intention, which is by its very nature once-and-for-all.”

This interpretation not only restores the similarity between judgments of desirability and intentions; it also helps to explain why Davidson is so worried about judging desirable a kind of action that may include undesirable instances. Whereas an unqualified or categorical judgment can be ventured on the basis of incomplete information, because it can still be reconsidered, a once-and-for-all judgment is closed to reconsideration, and so it had better be right the first time.

The Temporal Profile of Action

If this problem is the one that worries Davidson, then he has put his finger on a crucial but little noticed feature of practical reasoning as conceived by most philosophers. Philosophers conceive of practical reasoning as proceeding under a deadline. In order to cause action, the agent’s intention must precede it. If, alternatively, it consists in the action—a possibility that Davidson envisions—it must be cotemporaneous with it. And in either case, the intention will be an irrevocable conclusion, which will close practical reasoning once and for all. I think that this conception of practical reasoning is mistaken and that it has distorted philosophical views of how practical reasoning works.[5]

I will shortly present an alternative conception of practical reasoning according to which practical reasoning generally supervises behavior that is already underway, having been initiated without any reasoning at all. Before I introduce my conception of practical reasoning, I want to continue exploring how the temporal profile of action is typically conceived.

Consider this famous passage from Christine Korsgaard’s Sources of Normativity:[6]

I desire and I find myself with a powerful impulse to act. But I back up and bring that impulse into view and then I have a certain distance. Now the impulse doesn’t dominate me and now I have a problem. Shall I act? Is this desire really a reason to act? The reflective mind cannot settle for … a desire, not just as such. It needs a reason. Otherwise, at least as long as it reflects, it cannot go forward.

This passage describes a sequence of events: “then I have a certain distance,” “now I have a problem,” “as long as [the mind] reflects, it cannot go forward.” Action awaits the discovery—or, even more time-consuming, the construction—of a reason for acting.

I don’t say that Korsgaard intended this passage to be interpreted literally as the description of a temporal sequence. Maybe no philosopher of practical reasoning consciously envisions it as time-consuming. The fact remains that almost every philosopher of practical reasoning describes it in such terms, sometimes disavowing their implications but, I suspect, never effectively canceling them.

For another famous example, consider Bernard Williams’s description of a man who has “one thought too many”. The man is standing on the deck of a sinking ship, watching his wife thrash about in the water below. He wonders whether it is morally permissible to save his wife in preference to other victims:[7]

[T]he consideration that it was his wife is certainly, for instance, an explanation which should silence comment. But something more ambitious than this is usually intended, essentially involving the idea that moral principle can legitimate his preference, yielding the conclusion that in situations of this kind it is at least all right (morally permissible) to save one’s wife. ... But this construction provides the agent with one thought too many: it might have been hoped by some (for instance, by his wife) that his motivating thought, fully spelled out, would be the thought that it was his wife, not that it was his wife and that in situations of this kind it is permissible to save one’s wife.

Williams’s point is that the thought of moral permissibility would be “one too many” in the sense that, by thinking it, the man would betray less than complete devotion to his wife. That’s why his wife might hope that he wouldn’t think it. Unfortunately, Williams has couched this point in a story about an emergency, in which an agent must make a split-second choice. His description of the moral thought as one too many therefore steals plausibility from a different interpretation, namely, that the thought would be one more than the man has time to entertain before having to dive in. Under this interpretation, thoughts are discrete units, each of which takes time to entertain. Thinking one more or one less thought entails thinking for more or less time.

There is of course such a thing as overthinking a decision, of deliberating too much. But deliberation is not a mode of practical reasoning: it’s a procedure ancillary to practical reasoning, just as procedures of calculation are ancillary to arithmetic. Deliberation is like a checklist, a mental procedure that we use to help us think. It’s something we do. Indeed, it’s something we do deliberately—which cannot mean “on the basis of deliberation”, lest it lead to a vicious regress.[8]

Now, if the word ‘thinking’ is being used to mean deliberating, then the man in Williams’s story shouldn’t think at all: a single thought would already be too many, since he shouldn’t take time to deliberate. In this sense of the word, he should jump without a thought, though not of course without a reason.

If however ‘thinking’ is not being used for a mental procedure, then no assumptions can be made about how much or how little thinking would fit into the interval between a man’s look and his leap.[9] I’ve never stood on a burning deck, but I am sure that if I did, all sorts of things would cross my mind: the bilious green of the waves; the jaunty bobbing of lifeboats; screams from the swimming pool, of all places; the time I saw someone drown in the college pool; Kate Winslet. They say, in fact, that my whole life would pass before my eyes. How long would any of this mentation take? It’s not a serious question.

If a thought is not a step in a mental procedure, then it can be one too many only because it would be irrelevant, not because it would be too time-consuming. And in that case, the phrase “one too many” does not express an argument for the conclusion that thoughts of morality would be irrelevant; it just expresses the conclusion.

I don’t say that Williams intended the phrase to express an argument; I say only that the phrase is often read as expressing one. The argument depends on the same assumption that we found in Davidson, namely, that practical reasoning proceeds under a deadline, beyond which it is over once and for all: the man in Williams’s story has only so much time to think before acting, and so he has only so much time to think. To this assumption, the argument adds the further assumption that the requisite kind of thinking is a time-consuming procedure. According to this assumption, having only so much time to think entails having time for only so many thoughts.

Counting the Reasons

The conception of practical reasoning as a time-consuming procedure with a deadline is also implicit in the language that philosophers use to describe the normative force of reasons for acting. They say, for example, that reasons for an action are considerations that “count in favor” of it, and that nothing further can be added by way of explicating the concept of a reason or the concept of “counting in favor”. Counting is thus characterized as the unanalyzable essence of reasons for acting.

Unfortunately, the nature of practical reasoning is obscured rather than clarified by talk of counting. Counting is not in the first instance something that considerations do; it’s something that people do. Counting is an activity, and it is an essentially rule-governed activity. It is governed by rules for incrementing a sum by units corresponding to items of some kind, individuated in some way. When we speak of things rather than people as counting, we mean that those things are to be counted under such a set of rules. When the students ask, “Does spelling count?”, they are asking whether the rules for grading their work provide for points to be added or subtracted for words spelled correctly or incorrectly. When we say, “Close doesn’t count in horseshoes,” we’re talking about the rules for scoring the game.

The rules under which spelling counts on an assignment need not be fully determinate, but they cannot be utterly indeterminate either. If bad spelling merely has some probability of lowering a student’s grade by putting the teacher into a bad mood, then spelling doesn’t count; it makes a difference, but not by “counting”. In order for spelling to count, there has to be some rule-governed method for counting it, however vague or imprecise the rules may be.

To speak of considerations as counting in favor of an action therefore raises the question, How do they count? Whether interpreted literally or metaphorically, such talk suggests that there are rules, however vague or imprecise, for scoring or grading actions by enumerating pros and cons. So the phrase “counts in favor of” says significantly more than “is a reason for”, unless the notion of counting is explicitly canceled—in which case, nothing of the phrase is left.[10]

The problem, for present purposes, is that gratuitously introducing the notion of counting leaves the impression that reasons are essentially to be counted—hence, items in a time-consuming mental procedure that ultimately arrives at a bottom line. If reasons are considerations that count, then only so many of them can enter an agent’s practical reasoning, since only so many can be counted in time for him to act.

A similar confusion is introduced by the image of reasons as having weights. Here is an especially vivid exposition of that image by John Broome[11]:

Suppose you ought to Φ. An explanation strictly analogous to mechanical weighing would be this. There are reasons for you to Φ and reasons for you not to Φ. Each reason is associated with a number that represents its weight. The numbers associated with the reasons to Φ add up to more than the numbers associated with the reasons not to Φ. That is why you ought to Φ.

… [W]hen the fact that you ought to Φ is explained by pro tanto reasons, the explanation retains central elements of the mechanical analogy. It includes one or more reasons for you to Φ, and it may also include reasons for you not to Φ. These reasons are analogous to the objects in the left-hand and right-hand pans of the scales. Each reason is associated with a metaphorical weight. This weight need not be anything so precise as a number; it may be an entity of some vaguer sort. The reasons for you to Φ and those for you not to Φ are aggregated or weighed together in some way. The aggregate is some function of the weights of the individual reasons. The function may not be simply additive, as it is the mechanical case. It may be a complicated function, and the specific nature of the reasons may influence it. Finally, the aggregate comes out in favour of your Φ ing, and that is why you ought to Φ.

Strictly interpreted, Broome is speaking here of the weight that reasons have, not of any weighing procedure. He isn’t saying that practical reasoning is a process of weighing. But if the guiding force of reasons is analogous to weight, how can it guide an agent unless he weighs them? They have to be loaded into the pans of a scale and left untouched while the beam comes to rest. How else can they exert their normativity?

When we ask someone why he has done something, or what he proposes to do, the reply often takes the form of one or two considerations, discrete propositions that might be entertained at a stroke and counted or weighed on one or the other side of a question. The agent may even tick off these considerations, as if counting them, or mime the action of hefting one in each palm to compare their weights. But these are tropes of practical discourse, not windows into the mentation that actually leads to action.

The Supervisory Conception

I now want to present an alternative conception of the relation between practical reasoning and time. In my view, practical reasoning does not necessarily precede action, much less take time in advance.

In his Principles of Psychology, William James describes how we get ourselves out of a warm bed on a cold morning. “If I may generalize from my own experience,” he says, “we more often than not get up without any struggle or decision at all. We suddenly find that we have got up.”[12] Truer words were never written about practical reasoning. We think about getting up; that thought is swept away by many others that are irrelevant to the moment; and then we are already on our feet, without having been pushed by any occurrent thought.

Here I must insist on the empirical method: argumentation won’t do. Perform some careful observation of yourself over the course of the next few days. You cannot plan this observation immediately before acting, but neither can you leave it until after you have acted. Rather, just after you have begun to act—even as you are rising from your chair (or bed)—you must remember to observe yourself acting. Discard the cases in which you remembered to look before you started to rise. If you start to look too soon, you will just enact your prior conception of practical reasoning for the benefit of yourself as spectator. In order to see what actually happens when you act, you have to catch yourself, as it were, unawares. So the relevant cases are the ones in which you remember only after you have started to act, so that you weren’t watching at the crucial moment. In those cases, notice what passed through your mind immediately before you stirred. If I may generalize from my own experience (to echo James), you will find that if you had any thought at all of getting up, it was more of a speculation about getting up than a decision to do so. Then suddenly you were getting up.

The articulated mental accompaniments of action are not of the form: “I want a drink; there’s beer in the fridge; so I’ll get up and go to the fridge.” They are rather of the form: “A drink would be nice. Hey, who took the last beer?”

The fact that the intervening behavior didn’t follow any practical thinking does not entail that it was automatic or rote. There was practical thinking involved, but it was largely perceptual and supervisory, not deliberative. That is, you perceived yourself getting up and going to the kitchen, which (as you are aware) is the location of the beer, which (as you are aware) is what you had been thinking of fondly. These perceptions and implicit beliefs did not precede and initiate your behavior; they registered it and then followed along with it. Their role was to supervise the behavior by ensuring that it was something that made sense in the circumstances. It was your body that initiated the action; your conscious mind watched and thought, first, “I get it. OK then,” and thereafter, “So far, so good.” I defy you to find that anything else has gone through your mind—provided that you remember to look at just the right moment.

In my view, you don’t even individuate actions in advance, much less reason about them. You produce a continuous stream of behavior, which practical reasoning supervises and, in the course of supervising, places under action concepts.

As Davidson emphasized, most of the things you do, you do by doing other things or in the course of doing other things. You turn on the light by flipping a switch, which you do by flicking your finger, and so on; you turn on the light in the course of illuminating the room, which you do in the course of looking for your keys, and so on. Actions telescope. But philosophers generally assume that the telescope of actions is assembled in advance out of predefined segments, which are represented in an antecedent action plan. This assumption is what I deny. “Looking for keys, by illuminating room, by turning on light, by flipping switch, by flicking finger” is a post facto or perhaps in medio facto description of a undifferentiated stream of behavior. Each so-called action is abstracted from a fluid dance in which you also shorten your stride, grasp and release a knob, step and pause, sniff the air, peer into corners, turn your torso and your gaze, and so on. To be sure, you don’t do all of those things intentionally, whereas you intentionally flip the switch. But again: just after flipping a light switch, remember to ask yourself how that supposedly intentional action stood out from the rest of the dance. You will find that it didn’t, even if you were in fact looking for your keys.

Discrete act-descriptions get attached to behavior at various points in the process. Sometimes practical reason frames an act-description antecedently but needn’t hand it down as an order if behavior spontaneously rises to meet it. Sometimes practical reason frames an act-description simultaneously with the behavior it describes, so that reasoning and behavior proceed in step. Sometimes practical reason lags behind, and you find yourself standing barefoot on a cold floor, or thirsty at the door of the fridge, describing your behavior retrospectively. And sometimes, though not very often, practical reason has to take charge. In that case, you intend your actions in the classical sense, framing a plan and then carrying it out. Otherwise, brain and body go about their business, and your intending takes the form of superintending.

To be sure, my introspective experiment detects only what goes on in your conscious mind as you act; it doesn’t exclude the possibility that the balance of desire-belief reasons is guiding your behavior below the level of consciousness. And indeed, there is no denying that your behavior is often guided by desires and beliefs. The question is whether that guidance constitutes your behavior as action by exerting the normative force of reasons, as Davidson and others believe.

There are actions that have no desire or pro-attitude behind them, for example, expressive actions such as jumping for joy.[13] When my son called to say that he and his wife were going to have a baby, I really did jump for joy, and I was jumping out of joy itself, an emotion whose object was the prospect of a grandchild. I was not jumping out of a desire to jump, much less a desire to vent my joy by jumping.

There are also behaviors that are caused and rationalized by desires but do not clearly qualify as actions. Once when I was introduced to someone, I said, “Pleased to meet me”—an utterance motivated, I then realized, by a desire to tell my startled hearer how he should feel about the introduction and, by implication, to disavow the reciprocal feeling. I was surprised to hear the words come out of my mouth, and so I felt that, although mine was the mouth they came out of, it wasn’t me speaking. My speech act somehow fell short of action. It fell short because I spoke unwittingly, whereas the speech that amounts to action is spoken wittingly. Speech action consists in speech acts knowingly performed.

Note that my saying “Pleased to meet me” was not due to a mechanical failure in speech production: it wasn’t a slip of the tongue, like saying “Meased to pleet you.” Saying “Pleased to meet me” was a mechanical success at saying something that, at some level, I wanted to say. But my wanting to say it was not sufficient for constituting the utterance as an action with me as its agent. The sign that it was merely a speech act, not a speech action, is that it surprised me, as the speech acts of other speakers sometimes do.

In order to know what I’m saying, I have to say what I am already prepared to hear myself say, so that my saying it will not be a surprise. In the case of my embarrassing blooper, I was prepared to hear “Pleased to meet you” but ended up saying something else, for which I wasn’t prepared and by which I was therefore surprised. And behavior that surprises me, though motivated by desire and belief, does not count as my action.

I rarely say things for which I am not prepared, and that’s no accident. The reason is not that I usually prepare speeches before I make them. Preparing a speech is indeed a way of being prepared for it, but mostly I’m prepared for what I say because my next utterance is both prompted and prefigured by the situation, including my state of mind. The situation provokes the utterance; and on the basis of it I could, if asked, anticipate what sort of utterance it would provoke (though of course I’m not asked to anticipate it, and I don’t). I both say what comes next and implicitly know what comes next, so I’m not surprised by my saying it.

When I don’t know what comes next, I may have nothing to say, and, in any case, I couldn’t predict what I would say if I said anything. And even if I do have something to say, in the sense that there is something I would say if anything, I am usually inhibited from speaking if I feel ignorant as to what it will be. And that, in the last analysis, is why I am rarely surprised by what I hear myself say: I don’t open my mouth until I have the sense that I know, if only latently, what is going to come out. I must, then, be aiming to avoid surprises—to know what I’m saying, not as one of the ends or desired outcomes of my behavior but as the manner in which I behave, namely, to speak wittingly rather than unwittingly.

The sense of latently knowing what’s coming next in my behavior often arises from having followed it successfully until now. The word ‘follow’ here does not connote temporal succession. When we follow someone’s talk in a printed copy of his paper, we read along with him, simultaneously; we would only confuse ourselves if we followed temporally, by reading each word after he said it. I can usually follow my behavior, in the sense of simultaneously thinking along with it, because it is both prompted and prefigured by the same circumstances. And if I have been successfully thinking along with my behavior, then I gain confidence in being able to continue, and so I am no longer restrained by the aim of avoiding surprises. But if my behavior takes a surprising turn, the aim of avoiding surprises kicks in, and I stop doing anything until I have prepared myself for it by thinking ahead. And then I gradually regain my confidence in being able to follow along.

I would argue that this supervisory process is the final stage of practical reasoning, with which it is fully continuous. I cannot rehearse the whole argument here, but let me offer just one of the intuitions behind it.

When someone acts irrationally, we often say that he doesn’t know what he’s doing. In challenging him directly, we are just as likely to ask “What are you doing?” as to ask “Why are you doing that?” Behind these remarks is the assumption that a person acting rationally has a high-level, explanatory conception of what he is doing, a conception that sets his immediate behavior in the context of its outer circumstances and inner motives, which needn’t be desires and beliefs, since they may be, for example, emotions such as joy. (“What on earth are you doing?” my wife asked, covering the mouthpiece of the phone. If I hadn’t thought it was obvious, I would have explained: “Jumping for joy.”) Even if we ask someone why he is doing something, he may come back with an answer to the question what he is doing: “Why are you pacing the halls?” “I’m working on my dissertation.” These are the descriptions under which the person is following along with his behavior in the capacity of a supervisor.

To be sure, we usually have in mind some desired end to which our behavior will be conducive. But the conception of our behavior as directed at a desired end is a high-level, explanatory description under which we know what we’re doing. Our behavior is indeed desirable to us because we desire the end, and it is indeed guided by our knowledge of its efficacy in producing the end. But the desirability of our behavior is not the same as its rationality. To say that we usually act on a conception of our action as a means to a desired end is not yet to say how that conception rationalizes our action. Does it rationalize our action by showing it to be efficacious? by showing its effects to be desirable? or by giving us a grasp of what we are doing? Well, “jumping for joy” rationalizes an action without doing either of the first two. I say that rationalizing action consists in the third.

Of course, we sometimes come up with a means-end conception before we implement it, at a time when it doesn’t yet describe what we’re doing. But that conception does prepare us to see ourselves implementing it, so that we know what we’re doing when we do so. Our positive motivation for implementing the conception is our desire for the end, but that motive would be restrained if we didn’t feel prepared to see what we were about to do, or we were surprised when we began doing it. In addition to the positive motive, then, there is the conception under which the action is allowed to proceed by the supervisory process. In my view, the desire is our motive merely; our reason is the conception under which our behavior is supervised.

According to this supervisory conception, practical reasoning is not in the driver’s seat of action; it’s in the passenger’s seat. Better: practical reasoning is your spouse in the passenger’s seat, mostly following along silently as you drive; sometimes warning you of upcoming turns; sometimes telling you to take turns that you are already taking anyway, thank you very much; sometimes pointing out that you just took a wrong turn; and sometimes giving you just the direction you need, right when you need it.

Most of the time, you drive on automatic pilot, as we say. The automatic driver is your skilled, intelligent, goal-seeking mind, which can handle the car by itself most of the time. If you have a live pilot in the passenger seat, he plays the role of the human pilots in modern airliners, supervising the automated systems. If you are driving alone, then you play both roles, supervising your inner automatic driver. You supervise that automated system and intervene only if necessary.

The automatic piloting that is done by your intelligent, goal-seeking mind includes what Davidson identified as practical reasoning. That is, it includes desire-belief motivation. It also includes course correction in response to perceptual feedback about progress toward the desired outcome. (Davidson left that part out.) In my view, however, practical reasoning should be identified neither with desire-belief motivation nor with the attendant feedback and course correction. Those processes are what practical reasoning supervises; practical reasoning is the supervisor.

What argument do I have for this supervisory conception of practical reasoning? First answer me this: What argument is there for the standard, deliberative conception?

I have suggested several debunking explanations for the nearly universal belief in the standard conception—the literal interpretation of metaphors for the normative force of reasons, the conflation of practical reasoning with practical discourse or with its ancillary procedures. To these explanations I would add the sheer force of inertia. The standard conception of practical reasoning carries the weight of unquestioned lore, a weight that has stifled progress in the philosophy of action since Davidson, if not since Hume. The philosophy of action long ago entered the realm of fan fiction, a genre in which volunteer authors spin out the adventures of a character whom everyone knows well enough to distinguish between the fictionally consistent inventions about him and the fictionally inconsistent ones. The Rational Agent has taken his place alongside Harry Potter and Han Solo as a figure whose character is sufficiently determinate to give the ring of truth to some inventions about him but not others. Unfortunately, no reason has ever been given for believing that he actually exists.

My case for the supervisory conception of practical reasoning is not an argument but an introspective experiment, for which I have provided very specific instructions. Don’t watch yourself decide to do something: you’ll just see yourself enact the standard conception, in which you too have been enculturated. Rather, catch yourself when you have just started doing something, and cast a retrospective glance at the thinking that preceded your starting to do it. Where you expected to find the mythical Rational Agent, you’ll find the Supervisor instead.

If the supervisory conception of practical reasoning is correct, what follows?

One thing that doesn’t follow is any conclusion about moral responsibility. The moral responsibility of a navigator may be somewhat different from that of a driver, but so long as the navigator has or could have some influence on the route taken, he can certainly bear some responsibility for it.

A second implication of the supervisory conception is that automaticity in behavior is not the exception but the rule. Sub-agential mechanisms do not have to wrest control of our behavior away from a rational driver. There is no rational driver, only a rational passenger-navigator, whose task most of the time is to supervise behavior controlled by sub-agential mechanisms.

Finally, the supervisory conception implies that practical reasoning is not a time-consuming procedure that is carried out under a deadline. We don’t necessarily think before we act; usually, we think while we behave, and the combination ends up constituting action.


1. This chapter was written for a conference on time and agency sponsored by the George Washington University Department of Philosophy. It will appear in Roman Altshuler and Michael Sigrist, eds., Time and the Philosophy of Action (Oxford: Routledge, 2015).

2. Donald Davidson, “Actions, Reasons, and Causes,” in Essays on Actions and Events (Oxford: Clarendon Press, 1980), 6.

3. “Intending,” in ibid., 97.

4. Ibid., 99.

5. For another philosopher who questions this conception, see Talbot Brewer, The Retrieval of Ethics (Oxford: Oxford University Press, 2009), ch. 3.

6. Christine M. Korsgaard, The Sources of Normativity (Cambridge, UK: Cambridge University Press, 1996), 9.

7. “Persons, character and morality,” in Moral Luck: Philosophical Papers 1973-1980 (Cambridge: Cambridge University Press, 1981), 18.

8. This point is explored at length by Nomy Arpaly and Timothy Schroeder in “Deliberation and Acting for Reasons”, The Philosophical Review 121 (2012): 209–39.

9. See Gilbert Ryle, “Thinking and Reflecting,” Royal Institute of Philosophy Lectures 1 (1968): 216: “When we think in the abstract about thinking, it is usually reflecting, calculating, deliberating, etc. that we attend to. … Indeed it is just because reflecting is what we start off by considering, that we later on feel a strong pressure to suppose that for the tennis-player to be thinking what he is doing, he must be sandwiching some fleeting stretches of reflecting between some stretches of running, racquet-swinging, and ball-watching.”

10. There is a temptation to think that the counting rule for reasons is to add up weighted sums for and against an action. But if the weighted sums are sums of reasons, and reasons are considerations that count for or against an action, then the sense in which they count cannot be given by the rule of computing their weighted sum. The rule for counting reasons cannot be the rule of counting them according to their weight, if their weight is determined by how much they count. Such a counting rule would be vacuous. A substantive counting rule for reasons would have to be the rule of increasing the sum for each consideration that qualifies as a reason in a sense that’s independent of how much it counts. A different rule would therefore be needed to determine a reason’s weight. By itself, then, the notion of “counting in favor” appears to add nothing to the notion of reasons—except for the misleading suggestion of a time-consuming activity.

11. John Broome, “Reasons,” in Reason and Value; Themes from the Moral Philosophy of Joseph Raz, R. Jay Wallace, Philip Pettit, Samuel Scheffler, and Michael Smith, eds. (Oxford: Clarendon Press, 2004), 36–37. I’ve excised the second paragraph of this passage, which reads as follows: “Such a strictly analogous explanation rarely seems appropriate. For one thing, it often seems inappropriate to associate a reason with anything so precise as a number to represent its weight. Secondly, although we can aggregate together the weights of several reasons, to aggregate them simply by adding up also often seems inappropriate. So-called ‘organic’ interactions between reasons often mean that their aggregate effect differs from the total of their weights.” One might have thought that this concession completely undermines the analogy between normative force and weight, but Broome continues, in the third paragraph, as if nothing has been conceded.

12. The Principles of Psychology, vol. 2 (New York: Dover Publications, 1950), 524. Unfortunately, James goes on to speak of “an idea” that “flashes across us”, which in my experience rarely occurs.

13. See Rosalind Hursthouse, “Arational Actions,” Journal of Philosophy 88 (1991): 57–68.

 Originally published at http://dx.doi.org/10.3998/maize.13240734.0001.001



This article and its reviews are distributed under the terms of the Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and redistribution in any medium, provided that the original author and source are credited.