For much of his academic life Davidson tormented his critics with seemingly unanswerable arguments concerning an ancient problem in philosophy, a problem historically related to the question whether reason is or is not the slave of passion. Hume thought it was; Plato did not. The modern formulation of the central arguments related to this question belongs mainly to Davidson. The ancient formulation of the problem goes back to Plato who rejected the supremacy of the passions. His opponents he describes as follows:

They maintain that there are many who recognize the best but are unwilling to act on it. It may be open to them, but they do otherwise. Whenever I ask what can be the reason for this, they answer that those who act in this way are overcome by pleasure or pain or some other of the things I mentioned now. (Protagoras 352d-e)

The man overcome by pleasure is said to display a weakness of the will (akrasia). Davidson sees this question as part of a larger question, having to do not with right and wrong but, rather, having to do with the more general problem of explaining the nature of irrationality. There is something paradoxical about explaining irrationality. The difficulty he sees he succinctly describes.

The underlying paradox of irrationality, from which no theory can entirely escape, is this: if we explain it too well, we turn it into a concealed form of rationality; while if we assign incoherence too glibly, we merely compromise our ability to diagnose irrationality by withdrawing the background of rationality needed to justify any diagnosis at all. (Paradoxes of Irrationality. P. 303.

If I am right Davidson is mistaken in his understanding of the nature of the akratic’s irrationality. But before explaining why I want to get clear on how Davidson understands the problem. It is very important to keep in perspective the fact that Davidson sees the problem not as an ethical problem of knowingly doing the wrong thing, but rather as an action theoretical problem connected with the irrationality of doing something when one believes there is a better alternative. Consider, first, his definition of 'akrasia' or what he and others call ‘incontinence’:

D. In doing x an agent acts incontinently if and only if: (a) the agent does x intentionally; (b) the agent believes there is an alternative action y open to him; and (c) the agent judges that, all things considered, it would be better to do y than to do x. (A&E p. 22)

There is nothing obviously paradoxical about this. If there is a paradox it comes from accepting the following three propositions, all of which Davidson claims possess some sense in which they are all true.

P1. If an agent wants to x more than he wants to y and he believes himself free to either x or y, then he will intentionally do x if he does either x or y intentionally (op. cit. p. 23).

P2. If an agent judges that it would be better to do x than to do y, then he wants to do x more than he wants to do y (ibid.).

P3. There are incontinent actions.

What presents a paradox is maintaining any two without rejecting the third, for what we appear to have is an inconsistent triad. Davidson’s approach to escaping the paradox depends on maintaining a distinction between two kinds of judgments: a conditional judgment (‘all things considered y is better than x’ for example), and a non-conditional or categorical judgment (such as ‘y is better than x’), which elsewhere he calls an ‘all out’ judgment. There are two other components essential to Davidson’s treatment. First, there is the role of causation. This enters the picture inasmuch as the all things considered judgment serves as the causal antecedent of a categorical judgment which contradicts the categorical judgment that follows from the all things considered judgment, as ‘y is better than x’ follows from ‘all things considered y is better than x’. What occurs in the mind of the akratic is that the all things considered judgment that qualifies the agent by definition (D) as akratic literally causes him to affirm a categorical judgment that is not consistent with the consequent of that very same all things considered judgment, as ‘x is better than y’ contradicts ‘y is better than x’ arrived at from ‘all things considered y is better than x’. The convenience of viewing matters, thusly, is that P1-P2 are satisfied by the categorical judgment which is the effect of the all things considered judgment while the definition (D) is satisfied by the all things considered judgment. In this way it is possible for all three propositions to be true without an ensuing contradiction or inconsistency. There is one other important feature of Davidson’s approach that must be discussed before entertaining criticisms: what he calls The Principle of Continence. He gives us a number of formulations.

The best formulation in my opinion of the Principle of Continence goes this way:

…one should not intentionally perform an action that when one judges on the basis of what one deems to be all the available considerations that an alternative and accessible course of action would be better. (“Deception and Division” p. 139.

I am going to argue against Davidson's position in four ways. First, while Davidson emphasizes the importance of the fact that the akratic acts freely, his use of causation in explaining the possibility of weakness of will suggests that he may be violating his own requirement. Second,  I am going to maintain that Davidson is wrong in maintaining that the akratic takes a subset of reasons for believing that all things considered he ought to do one thing and applies them as justification for doing something else. I'll call this the "subset" argument. Third, I will argue that the akratic may act, not on reasons for doing one thing rather than another but on the basis of reasons for doing something with no consideration of reasons for doing another thing whatsoever. The akratic, moreover, may simply act on the basis of the property of an individual involved in the considered action, paying no attention for evidence in support of one of two disjunctive possibilities. Fourth, I will argue that the proper characterization of the type of irrationality exemplified by the akratic is to say that it is an instance of what has come to be known as "Moore's Paradox."

It is important to keep in mind that according to Davidson the issue involves conflict, in particular a conflict between alternative actions. The incontinent individual is caused to accept a belief that one is better than another contrary to his best judgment, all things considered. This affects the logic of the situation and provides the groundwork for my first objection to Davidson’s view of the problem.

If the incontinent individual is caused to accept a certain belief then actions based on that belief are free only in an attenuated sense. The individual is caused to behave a certain way against his better judgment and insofar as he acts not in accordance with his reasoning but, rather, as a consequence of effects over which he has no control it can be argued that he is not free, and insofar as he is not free he is not an agent, and inasmuch as he is not an agent he is not culpable for his actions. It is essential that the incontinent man, the akratic, be an agent, that he act intentionally, and in this instance the cause of his belief is not a reason. This is not, however, necessarily a fatal objection; nevertheless, attending carefully to the text one can only wonder why Davidson appears to shift back and forth between considering the akratic one who reasons badly, and one who is caused to act irrationally. I believe what Davidson might say is this: the akratic is caused to believe ‘a>b’ by the judgment that all things considered b>a. He is induced to come up with reasons for believing ‘a>b’ and it is on the basis of these reasons that he acts, not that he is caused to act by the all things considered judgment alone. There is, however, a more powerful objection to be made based on the fact that Davidson views the relation of (D) to P1 and P2 as essentially involving conflict, conflict the outcome of which is expressed in terms of comparative judgments about better and worse alternative actions.

It is best if we take an example of weakness of the will and pursue it as far as we can, until we begin to see the weaknesses, if there are any, in Davidson’s position. By doing so, we shall discover at the very least that Davidson's characterization of the akratic is insufficient to capture all bona fide cases of weakness of the will. Let’s consider the case where I am faced with the option of either feeding my dog or feeding a hungry stranger. Suppose I judge that all things considered it is better to feed the stranger than my dog; but suppose, as well, that even though I believe this is the right thing to do I opt to feed my dog. I do so intentionally - believing I am free to do either. What Davidson asks us to believe is that I am caused by my very judgment that all things considered it is better to feed the man to believe that it is better to feed the dog. I find this difficult to believe, but not logically nonsensical; so we proceed to consider that he is also telling us that on the basis of this belief I select a subset of reasons I used in judging that it is all things considered better to feed the man than the dog and apply them in arriving at the belief that it is better to feed the dog than the man. I find it difficult to believe that there is a single reason for feeding the man rather than the dog that could be used to justify feeding the dog instead of the man, but let’s allow this. What is Davidson's way out? I can think of one possibility with some backing in what he actually says in his reply to Grice and Baker. Davidson admits that ‘all things considered’ is ambiguous, and he argues that variations in its meaning simply determine different sense of ‘irrational’ (“Reply to Paul Grice and Judith Baker” p. 205). Assuming this to be his position, we must judge it weak. What Davidson wants to say is that ‘all things considered’ is to include not merely reasons but all the facts that go into stating those reasons. Here is where I think matter must go awry.

Davidson repeatedly claims (A&E. p. 40, “Reply to Paul Grice and Judith Baker” p. 204, “Deception and Division” p. 139) that the akratic acts on the basis of a subset of reasons which he has applied in arriving at his all things considered judgment, a judgment whose categorical entailment is comparative: ‘doing x is better than doing y’. If I list my reasons for feeding the dog and not the man, such as that I like the dog better, or that the dog has been loyal, whereas the man I don’t even know, I have not listed a single reason for doing feeding the man rather than the dog. This much I've said already. Now I may give as a reason for feeding the dog instead a noncomparative fact about the situation taken in its entirety, and not just a reason for judging the comparatives in arriving at the best possible action; viz, the fact that she is not human. But this hardly seems like a reason that would be included as a reason for feeding the man rather than the dog; but even if it were, the claim is that the set of reasons for dog are included in those for feeding the man rather than the pooch, and this is simply and obviously false. What would have to be smuggled in is a fact supporting neither alternative action, but just a fact about one of the subjects involved.

My point here is that Davidson would most likely take the position that we include in the extension of ‘all things considered’ all that goes into our reasons for judging that all things considered it is better to feed the man than the dog. Notice that it is a requirement and not a mere contingency that the reasons for judging that it is better to feed the dog than the man be included as a subset of the reasons for judging it better to feed the dog than the man. But is it actually the case that an agent, an akratic one at that, must reason in such a way, not merely that he might? If Davidson’s proposal is a solution to the apparent paradox it seems the agent must reason in such a way. 

Davidson might be thought to be maintaining that ‘all things considered’ includes considering reasons for judging ‘a>b’ in addition to those for judging ‘b>a’ in judging ‘b>a’. This would be a very awkward way of reasoning since we rarely include reasons for believing ‘not-p’ among our reasons for believing ‘p’, but if as - I believe - David Lewis once said anything can cause anything then the all things considered judgment may cause the agent to consider reasons for the opposite of a conclusion as constitutive of ones reasons ‘all things considered’. Let us consider Davidson's approach from a slightly different angle.

Davidson’s alleged paradox related to akrasia requires that one alternative possible action be pitted against another: I believe one is better, but I perform the other, instead. Without this competitive situation the comparative ‘better’ cannot be introduced in P1-P3. An alternative to Davidson is the claim that the akratic may, all things considered, believe that doing x is the right thing to do, but not do it, where no alternative course of action is under consideration. Kenny has pointed out that the specificity of the description of an action can be relevant to whether an action is voluntary. The following description:

Throwing the cargo overboard

may not be the appropriate description of an action deemed voluntary, whereas,

Throwing the cargo overboard to save oneself and one’s crew

may be appropriate. Kenny remarks:

Since both descriptions are truly applicable to the master’s action, it is ‘mixed’: it can be described either as wanted or unwanted. But when we are classifying actions as voluntary it is the more complete description that is appropriate, the description including the relevant circumstances. So, all things considered, the master’s action is voluntary. (A. Kenny in Aristotle’s Theory of the Will. Duckworth. 1979. p. 32).

I think this opens up the possibility of acting incontinently after taking into account “all things considered” where nothing concerning another possible alternative course of action is considered. Consider, then, Davidson’s Principle of Total Evidence. If I do consider an alternative, I may consider reasons for doing each, but if descriptive specificity increases in the case of one, y, reasons for doing it would still have to be included, according to Davidson, as a subset of reasons for doing x, where y is the incontinent action. But is it plausible to believe that as a description for doing y becomes increasingly detailed, that we can expect reasons for doing it under that detailed description to be considered among reasons for doing x? As our description becomes more and more detailed there is less reason to believe reasons for it will be included in the other. True we may consider whatever we like, but what is important is that I can voluntary do what all things considered I believe to be the wrong thing to do without considering the reasons for some alternative action; it may be argued that all I need consider is what would be sufficient to make that action voluntary.

Even if it were possible to argue successfully that ‘all thing considered’ judgments that ‘b>a’ include consideration of reasons for believing ‘a>b’ we would still have a problem with the fact that incontinence is only defined only over comparative judgments. When we speak of one event being more probable than another, we mean simply enough that the probability of one event has one numerical value and the probability of the other has a different one and one is greater than the other. The continuum decides the ordering relation in judging comparative probability. This is not the case with all comparatives, especially those closest to Davidson’s heart, ‘better’ and ‘best’. Herein lies the problem. It is only on the basis of an ordering of properties that appeal to the properties of actions can decide which action is better. Even so, there is no good reason for believing that the akratic must judge a comparative, such as that it is better to feed the man and not the dog rather than feed the dog rather than the man. He may judge that it is wrong to feed the dog but act on the basis of the color of the dog’s eyes or its kindly face. If one asks why he should do this, one might ask of Davidson why making a certain judgment causes a person to come up with reasons for doing the opposite. Such considerations lead to an altered view of what it is to be akratic. 

On this revised view, the akratic has reasons for believing a comparative is true, viz. that one course of action is better than another but he acts on the basis of belief unrelated to comparative evaluation, such as that some action has a certain property - one which is non-comparative. To take the case of the dog and the man, the agent judges that it is better to feed the man than the dog but acts on the basis of the dog’s having a certain property, such as a kindly face. If this is true, then Davidson’s definition of incontinence is insufficient to capture all instances of akrasia and is, therefore, insufficient, since he is acting not on the basis of reasons for preferring one action to another, but on the basis, rather, of acting on a single action without regard for how that basis figures in an evaluative judgment of a comparative nature. I will next entertain the Davidson’s discussion of how the irrationality of the akratic and the self-deceptive individual are related.

I want to do two things in the brief discussion that follows concerning self-deception and the irrationality of the akratic. First, I want to argue extend my remarks on Davidson’s view of the role of causation in self-deception, and second, I would like to introduce the possibility that the irrationality of both the self-deceptive individual and the akratic is more closely related to Moore’s paradox than what Davidson calls the paradox of irrationality. The quickest way to see the connection, according to Davidson, between akrasia and self-deception is to substitute ‘b>a’ for ‘p’ and ‘a>b’ for ‘not-p’ in the following description of self-deception in Davidson:

An agent is self-deceptive with respect to the proposition p under the following conditions: A has evidence on the basis of which he believes that p is more apt to be true than its negation; the thought that p, or the thought that he ought rationally to believe p, motivates A to act in such a way as to cause himself to believe the negation of p… All that self deception demands of the action is that the motive originate in a belief that p is true…and that the action be done with the intention of producing a belief in the negation of p. Finally, and it is this that makes self-deception a problem, the state that motivates self-deception and the state it produces coexist; in the strongest case, the belief that p not only causes a belief in the negation of p, but sustains it. (“Deception and Division” p. 145).

I want to challenge the role Davidson has causation play in both self-deception and akrasia. I begin with Davidson’s introduction of causation in his treatment of self-deception. Consider, then, four sentences.

1. D believes that he is bald.

2. D believes that he is not bald.

3. D believes that (he is bald and he is not bald).

4. D does not believe that he is bald.

Davidson contends that the belief as reported in (1) in the self-deceptive individual serves as cause of the belief reported in (2). He avers that while it may appear at first that (2) entails (4) but, he says, this would contradict (1) and so we must reject such an entailment. My problem with this begins with the fact that I’d sooner believe that what is reported in (1) does not cause (2) than believe that it is possible to believe I’m not bald without its being the case that I don’t believe I’m bald. Davidson’s dilemma is that he needs to explain how one arrives at inconsistency (in some sense) and since he can’t explain this in terms of reasoning he interjects causation. Is there a better alternative? I think there is. The alternative is to believe that the self-deceptive individual never gets to the belief reported in (1); that he has all the evidence, but will not make the obvious inference to the conclusion that he is bald. Davidson introduces the “unnecessary shuffle” because he sees this as a way of subsuming two problems, akrasia and self-deception, under a common category of irrationality. Is there an alternative? I think there is.

The sort of irrationality that the self-deceptive individual and the akratic have in common is that both are can be expressed as instantiations of Moore’s paradox. Moore’s paradox involves explaining rationally how the following assertion can be made sense of without contradiction: I am such and such but I don’t believe I’m such and such. My position is that for the akratic the relevant sentence becomes:

All things considered I believe b>a but I don’t believe b>a

Whereas for the self-deceptive individual the relevant sentence is:

All things considered I am bald, but I don’t believe I am bald.

This last sentence might be uttered by an individual, for example, intent on confessing that he is self-deceptive. If we accept this interpretation of the facts there is no need to introduce causation and yet we remain in a position to lump the akratic together with the self deceptive individual. In addition, we avoid the counterintuitive claim that from ‘I believe that not p’ does not entail ‘I don’t believe p’.

 

^* Comments on Moore's Paradox and its application evolved out of discussion with Ray Perkins Jr.