Friday, January 25, 2008

Velleman on Action and Knowledge

In The Possibility of Practical Reason (the book, though not the article of that name), David Velleman argues that self-knowledge is a constitutive aim of action. He develops his argument neatly by comparing action, or full-blooded action from such mere activity, or defective action, as that performed in a Freudian slip. He recalls Freud’s example of the President of the Lower House of the Austrian Parliament, who opens the assembly with the declaration that it is closed. A performance of this sort (one that can be explained by the agent’s desires—not fully in this case, since the President knows he cannot close the assembly simply by declaring it closed, but Velleman has another Freudian example he presents to deal with that slight problem; I am dealing with this one for its relative simplicity), an act that occurs as a slip, Velleman takes it, is an example of something that is not fully an action. It is more than simply a piece of behavior, but it is not a full-blooded action, or an autonomous action. And Velleman argues that what makes it defective as an action is that the agent does not know what he is doing before he does it. Or, rather, what he does is not what he knowingly chooses to do. Thus Velleman advances self-knowledge (or, at least, knowledge of what one intends to do) as a constitutive aim of all action: action aims at being known by the agent before it is performed, and so an autonomous action is one that the agent knowingly chose to perform beforehand.

I want to question this account, but at first glance it seems reasonable: it seems like autonomy, or full-blooded agency, requires control. But it is not clear how one could have control over an action that one performs without first knowing what it is one is planning to do. Thus, insofar as full-blooded agency requires control, it might seem reasonable to think that self-knowledge is indeed a constitutive aim of such action (not, in other words, an explicit aim of the action, but an aim that actions must have in order to qualify as actions at all). It seems, however, that one feature of this account is that there is no cap on the knowledge required for autonomous action: the more knowledge (or pre-knowledge, since what is involved is knowledge of the action prior to its occurrence), the better. The President knows approximately what he is going to say, but declares the meeting closed instead of open. It thus might seem like he had some vague idea of what it was he meant to say, but had no rehearsed the actual words in his head. If he had—better yet, if he had also rehearsed the tone with which he was going to utter them—he action would have been more autonomous. What I want to suggest, however, is that this requirement is too strict: we do not normally require that an agent have pre-knowledge of his action in order to hold him morally responsible; nor do we generally need to have pre-knowledge in order to experience an action as in our control.

Why do we care about knowing what we are going to say before we say it? One reason is that we want to be understood by others, to communicate our thoughts—being understood is, almost certainly, a constitutive aim of normal speech. But this does not require that we choose our words before uttering them: we make ourselves understood in dialogue (and this is one way in which professional writing differs from everyday speech) by choosing our words as we say them. We do not normally need a prior process of choosing those words within an internal monologue, which in fact serves to take them out of the dialogical situation. What does one practice before an interview? Reciting a pre-written monologue about one’s research and one’s abilities, or being able to extemporaneously recite such monologues in various forms and in response to external cues? To be sure, there are situations where perfect control of what we say is called for—for example, performatives, like that spoken by Freud’s President, require precision. But these are certainly not normal cases of speech. What is normally required is not that we know exactly what we are going to say—imagine conversations like that!—but that we avoid making grossly inappropriate blunders. And pre-knowledge seems like an overly strict requirement to impose for such a goal.

Imagine the situations where you carefully choose the words you are going to say before saying them (similarly—actions that you thoroughly rehearse in your mind before performing them). What normally happens is that these words, when actually spoken, come out sounding artificial, rehearsed. Of course this isn’t always the case: actors are people who can make prepared speeches sound natural. But acting is obviously an exceptional situation, not the norm. My point is not that we don’t normally think through our words fully before speaking them. Velleman agrees with this, and simply responds that in such situations the constitutive aim (knowing what we are saying) has been “scaled back”; in any case, this would show only that most of our actions are not full-blooded actions, and that wouldn’t mean that the full-blooded actions aren’t the norm-setting ones.

So my point in noting that acting is an exceptional situation is not simply to point out that it is not normal, but rather to point out that it is abnormal. Acting—speaking words that one has already rehearsed—cannot be taken as a paradigm case of action for the simple reason that it is imitated action, or action intended to mimic action. Just as a parody cannot be taken as paradigmatic of the original (“Austin Powers” and “Our Man Flint” are not paradigmatic Secret Agent movies), so acted action is not paradigmatic of action as such. Insofar as rehearsal is involved in the production of an action, that action is excluded from being paradigmatic. And this is not because we do not typically rehearse actions, but because rehearsed action is not the sort of action we want to understand if we are trying to say something about human action as such. And action that is not paradigmatic, it seems to me, cannot reasonably be taken as norm-setting for action as such.

There are, of course, cases of rehearsal that are less theatrical: we have to practice certain movements in order to learn to perform them expertly. A martial-artist must practice a form in order to perform it flawlessly, for example. But a martial-art expert acts without rehearsing the move in his head. The point of the practice, of the rehearsal, is precisely that when he acts, he does not need once again go through his knowledge of the movements. This—as Hubert Dreyfus has argued—is the mark of expertise. And insofar as we are all—to some extent—expert agents, our agency likewise does not require pre-knowledge to its success.

Continue Reading...

Wednesday, January 9, 2008

Causal Theories of Action (II): The Limits of Explanation Abstractions

In the last post I looked at multiple reason scenarios (MRSs), cases where an agent has more than one reason to act and ends up acting for only one of those reasons. I looked at attempts by both causalists and non-causalists to use MRSs to argue for their approaches, and suggested that MRSs cannot provide a good foundation for any argument whatsoever unless one can dismiss the possibility that there are no such scenarios.

So how can one dismiss the possibility? How can one show that, at least sometimes, agents in possession of multiple reasons to A end up A-ing for only one of those reasons? Searle seems to appeal to this simply as a fact of experience: we just know that sometimes we act for a particular reason and not another. If so, Searle’s experience must be exceptionally shallow; it is a fairly common fact that, often, looking back upon our past actions we realize that we did not take those actions for the reasons we thought we did. In any case, introspection is notoriously unreliable in such cases, as any decent account of self-deception (my favorite, of course, is Sartre’s) will show. Take the example Mele gives: Al has two reasons to mow his lawn this morning (because it is convenient, and to spite his neighbor), but only mows the lawn for one of those reasons. Can Al be sure that he mowed the lawn for only one of those reasons through introspection? Al may well believe that he is mowing the lawn because this is a convenient time to mow the lawn and not because he wants to get back at his neighbor. But the fact that Al believes that he is not petty does not prove that he is really not petty. My point is not that introspection is fallible; everybody already knows that. My point is that introspection has limits, and this is one of them: introspection cannot tell us, in any situation, which reason, out of a set, we really acted on.

Searle’s entire introspective scenario, in fact, strikes me as wildly implausible. He insists that we simply know which reason we acted on, and that we know it because we chose that reason. But I can’t imagine this happening: if I have two competing reasons for doing something, and I end up doing it, there is only one type of case in which I can then be sure that I acted on only one of the reasons: if I experience one reason as motivationally stronger than the other. But this is precisely not something that I chose—what I chose, or at least perhaps experience myself as having chosen—was to act. But choosing to act is not the same as choosing a particular reason to be the reason I act on. And of the latter sort of choice, I doubt anyone can have any experience whatsoever. Reflecting on such experiences, instead, seems to support the idea that our reasons act on us and not the other way around.

Can Mele or Davidson do better? I don’t think so. For how would they go about justifying the claim that, sometimes, agents with multiple reasons for A-ing actually A for one particular reason? Well, they could appeal to the causal theory: since every action is caused by some reason, it is at least theoretically possible to be certain that a particular reason, and not any other, is responsible for some action. But of course this would be question-begging: we cannot use the conclusion that the MRS is supposed to yield in order to justify the plausibility of the MRS. Davidson, in discussing a somewhat different point (responding to the claim that knowledge of causal laws is always inductive), tells us the following:

You may be wrong about which motive made you do it. The fact that you may be wrong does not show that in general it makes sense to ask you how you know what your reasons were or to ask for your evidence… Then your knowledge of your own reasons for your actions is not generally inductive, for where there is induction, there is evidence. Does this show the knowledge is not causal? I cannot see that it does… In any case, in order to know that a singular causal statement is true, it is not necessary to know the truth of a law; it is necessary only to know that some law covering the events at hand exists.(Davidson, 18)

Not only might we be wrong about our reasons for an action, but we also have no evidence to appeal to. In other words, there is no reason to think that we aren’t always wrong; there is no reason to think that we have any causal knowledge whatsoever when it comes to our reasons. We might know that, in general, some of our reasons are causally involved in producing our actions, but we have no knowledge about what reasons in particular might be involved. And we can know the former point only because we “know that some law covering the events at hand exists”; that is, we know that events are caused. But, again, this does not show that some particular reason, rather than all available reasons, is involved in the causal production of any action.

In fact, the example Mele uses is itself pretty strained: I am not at all sure it is possible for Al to mow the lawn only because it is convenient, and not because he wants to get back at his neighbor. The problem is this: if one reason Al has to mow the lawn this morning is to get back at his neighbor, then he is aware that his mowing the lawn this morning will irritate his neighbor. And so it seems that if Al does mow his lawn, he does so with the awareness that he is irritating his neighbor; and unless he specifically wants to irritate his neighbor by his action, then the reason is—if anything—a reason against his mowing the lawn, not for it. Of course there is a way out: Al might have both reasons but act on only one of them if he simply forgets about the other reason. This is perfectly feasible. But in such a situation the one reason he remembers is active; the other reason is not, but only because it is not functioning as a reason at all. And this comes down to saying that the MRSs can only work properly by cheating, that is, by treating all but one of the existing reasons for an action as simply inert. And this seems much like the scenario where the agent really has only one reason for action to start with.

So, getting to the point: It should be fairly clear that causal theories of action, in their exposition, are generally abstractions. In real life, we do not act for precisely one reason unless we are only aware of one reason. More commonly, I suspect, our actions are produced by a causal circumstance rather than a given cause. The causal circumstance might be quite broad. Let’s say that I decide to eat a cake. My reasons for this may well include not only that the cake looks good to me, but also a general predisposition to sweets, or a gnawing hunger and lack of hot dogs in the vicinity, a general proattitude toward eating based on the past experience of its feeling good, perhaps also a commitment to keeping my strength up or not going too long between meals so as to avoid putting on weight. Each of these commitments, in turn, is likely to have other commitments behind it. To speak of an action being caused by one reason is, generally, to oversimplify matters—not just because the causes involved go far beyond any one reason, but because the relevant causes involved, the reasons, are not usually limited to one. It is not only philosophers who oversimplify, of course. In everyday life, if we stop to think about it, we often imagine that there is a particular reason for our actions. Having such a reason—or a defined set of reasons—comes in handy when we need to justify an action to others, or even to ourselves. But this focus on particular reasons may well serve to cover up a deeper fact: that our actions are, more commonly, caused by a broad range of dispositions, desires, beliefs, habits, urges, and commitments. Focusing on particular causes rather than on the whole relevant causal circumstance—what might be called “the will” in the proper sense—has benefits. It allows us to individuate the causes of actions in order to find the ones that best explain that action. But it has the disadvantage of making us forget that the range of relevant causes of any action is, normally, far wider than often thought. Understanding how these causes can operate together in producing an action—as a unified will—is thus a project underlying any causal theory of action.

The suggestion, then, is that there is a gap between explanation and reality. We rely on abstractions in explanations because we require intelligibility. And because are minds have relatively simple computational abilities, to have intelligibility we need simple models. But reality is not simple; action does not, generally, issue from one particular reason, even if one particular reason provides the most effective explanation of that action. Sticking with the Davidsonian account, then, has a problematic disadvantage: what it gives us is a theory of action-explanation. But it says little, if anything at all, about action.

Davidson, Donald. (1963) “Actions, Reasons, and Causes” in Essays on Actions & Events, 3-20.

Continue Reading...