People sometimes knowingly and intentionally act contrary to their own best judgment, which seems counter-intuitive since it challenges a self-evident doctrine: “that, in so far as a person acts intentionally he acts in the light of what he imagines (judges) to be the better [option]” (Davidson 2001, p. 22). Acting intentionally in light of what one imagines is a better or most optimal option seems to be self-evident because it is impossible to imagine why anyone would not perform some action that they judged to be better.
Acting contrary to one’s own best judgment is called “weakness of will.” Donald Davidson has offered one account of how weakness of will is possible. Others, of course, have offered their own account for or against the possibility of weakness of will. For other accounts see chapter 5 “Backsliding” of R.M. Hare's Freedom and Reason (Oxford University Press, 1965, p. 67-85) and contributions to Geoffrey Mortimore's Weakness of Will (London: St. Martin’s Press, 1971).
Davidson has characterized an action that amounts to weakness of will in the following way:
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. (Davidson 2001, p. 22)
Davidson believes the problem of incontinence arises from the conjunction of two principles, which connect judgment, motivation, and action, with the belief that incontinence exists. He spells out the two principles in this way:
P1. If an agent wants to do x more than he wants to do y and he believes himself free to do either x or y, then he will intentionally do x if he does either x or y intentionally.
P2. If an agent judges that it would be better to do x than to do y, the he wants to do x more than he wants to do y. (Davidson 2001, p. 23)
What P1 and P2 entail is that if an agent judges it would be better to do x than to do y, and he believes that he is free to do either x or y, then the agent will intentionally do x if he does either x or y intentionally. The problem is that if we add (P3) An agent acts intentionally, counter to his own best judgment, P1-P3 form an inconsistent triad. These three principles comprise Davidson’s view of how the problem arises. Others may describe the possibility of weakness of will differently, or they may be inclined to say that weakness of will is impossible.
This post will not try to solve the problem of incontinence by eliminating one of its principles, P1-P3, which is a common way of resolving the problem of incontinence. For instance, see C.C.W. Taylor, “Plato, Hare and Davidson on Akrasia,” Mind 1980 (89): 499-518. Taylor explains that at least one of the two principles is false. The most likely candidate according to Taylor is Davidson’s P2 principle. Taylor fails to see the difference between P1 and P2 that Schueler and I do. Rather, my aim will be to explain Davidson’s position as clearly as possible and to acknowledge some modal issues in the problem that may eventually lead to its resolution. First, I will explain Davidson’s account of how weakness of the will is possible. Second, I will try to meet some of the common objections to Davidson’s account. Finally, I will conclude that P1-P3 are not incompatible, but that even if they are, such incompatibility does not necessarily rule out incontinent actions.
Davidson's Explanation of the Possibility of Weakness of Will
Let us start by explaining Davidson’s view of weakness of will. Davidson uses an analogy between evaluative and probabilistic statements to differentiate conditional from unconditional judgments. Doing this entails that an interpretation of the principles P1-P3 are consistent, and they ought to be accepted. For an agent who does y rather than x may judge x is better than y, but the judgment is a conditional judgment, all things considered. Many ethicists would not say “x is better than y, all things considered” is a conditional judgment. They also would be inclined to reject Davidson’s distinction between “all things considered” judgments and “sans phrase” judgments. According to Davidson, to judge that x is better than y, a preferential evaluative judgment should be the cause of some intention to do either x or y, but he proposes to distinguish between preferential evaluative judgments that carry with them an intention and those that do not. Davidson, thus, employs an analogy between evaluative statements and probabilistic statements to illuminate the logical formalism of prima facie judgments. Davidson believes we are not entitled to reason from probabilistic evidence, such as:
M1 If the barometer falls, it almost certainly will rain.
m1 The barometer is falling.
via modus ponens,to the conclusion:
C1 It almost certainly will rain. (Davidson 2001, p. 37)
Instead, we are merely entitled to say, “that the barometer is falling renders it probable that it will rain.” C1 differs from this proposition because C1 is a categorical assertion that follows logically from M1 and m1 whereas Davidson’s alternative proposition is not a categorical assertion following from M1 and m1 since he believes the logical form of M1 is not ‘p→q’. We cannot infer C1 from the conjunction of M1 and m1, even if it is suitably qualified by ‘almost certainly’ (Davidson 2001, p. 38).
When we consider an instance of probabilistic reasoning to a contrary conclusion, we understand why we cannot categorically assert C1. For example:
M2 Red skies at night, it almost certainly will not rain.
m2 The sky is red tonight.
C2 It almost certainly will not rain. (Davidson 2001, p. 37)
We can suppose circumstances in which both sets of premises are true that make the arguments valid; however, there is good reason to think that C1 and C2cannot both be true. Neither C2 would entail C1, nor vice versa. For C1 to follow from C2, almost certainly not-rain must necessarily entail not-(almost certainly rain). C2 entails not-C1; however, the conjunction of C1 and C2 is a contradiction. Thus, C1 and C2 are contraries.
Showing that C1 and C2 are contraries generates logical trouble for the formalization of probabilistic reasoning. The probabilistic evidence in the two syllogisms may support a prediction to two different states of affairs. In this case, the predictions are contraries. The syllogistic form cannot capture the form of probabilistic reasoning. Therefore, such reasoning shows that the premises do not logically entail the conclusion, and the conclusions must not have the form of C1 and C2.
Davidson’s way out of this struggle is to see matters in terms of ‘almost certainly’ as modifying the connective rather than modifying the conclusion. We usually use connectives to build longer sentences out of two or more shorter ones. Davidson uses connective in a similar fashion. Using ‘pr’ as a connective enables Davidson to construct a schema where the probabilistic predictions of contrary states of affairs seem to be possible under the circumstances. Probabilistic predictions should be schematized in the following way:
M3 pr(M1)
m3 (m1)
C3 pr(it will rain, M3 • m3)
M4 pr(M2)
m4 (m2)
C4pr (it will not rain, M4 • m4)
So, a part of the probabilistic evidence supports C3 and C4, and they are logically compatible.
Just as probabilistic reasoning may lead to contrary conclusions, so too can evaluativejudgments lead to practical conclusions that are contraries. Davidson believes something that is prima facie wrong cannot be treated as a “universally quantified conditional, but should be recognized to mean something like” a connective (Davidson 2001, p. 37). For instance, the proposition ‘Lying is prima facie wrong’ should be ‘That an act is a lie prima facie makes it wrong’. Similar to probabilistic reasoning, using ‘pf’ as a connective enables Davidson to construct a schema where the prima facie predictions of contrary states of affairs seems possible under the circumstances. Practical conclusions ought to be schematized in the following way:
M5 pf (x is better than y, x is an act of lying and y is refraining from an act of lying)
m5 A is an act of lying and B is refraining from lying
C5 pf (B is better than A, M5 • m5)
M6 pf (x is better than y, x is an act of mercy and y is refraining from an act of mercy)
m6 A is an act of mercy and B is refraining from an act of mercy
C6pf (A is better than B, M6 • m6) (Davidson 2001, p. 38)
With this formalism, we avoid logical difficulties that allow us to recognize considerations both for and against performing a given action. Both C5 and C6 can be true. Practical conclusions are not categorical; rather, they are prima facie judgments.
Neither probabilistic reasoning nor practical judgments are confined to considering one or two pieces of evidence. On the contrary, a prediction or a decision to act based on all the relevant information will be a better prediction or decision than that based on merely one piece or even some pieces of evidence. We aim for a prediction, such as pr(rain, e) or pr(not-rain, e), where ‘e’ represents all the relevant information, or a practical judgment, such as pf(a is better than b, e) or pf(b is better than a, e), where e represents all the believed relevant considerations,since these will turn out to be better predictions or decisions than ones based on quantitatively less information. Despite the fact that a prediction or decision based on all the relevant information is a better prediction or decision, it does not turn a statement into an unconditional one. All things considered judgments and unconditional judgments are distinguishable since all things considered judgments consider all of the agent’s beliefs while unconditional judgments require just one reason for an agent to make a prediction or decision. Unconditional judgments are detached from the premises, including all relevant information, that lend them support. Davidson’s formalism enables us to express conflicting considerations without contradiction, but his formalism does not enable us to resolve the problem of incontinence. Davidson believes that it is precisely the characteristic of prima facie judgments that enables them to express conflict without contradiction that precludes those judgments from being relevant to the execution of an all things considered judgment.
The unconditional judgment, or practical judgment sans phrase, is crucial to a resolution of the problem of incontinence. The distinction between conditional and unconditional judgments captures the difference between concluding that an agent ought to perform some action and an agent actually executing that conclusion. Unconditional practical judgments are ones detached from the premises that lend them support. Davidson says, “intentional action, I have argued in defending P1 and P2, is geared directly to unconditional judgments like ‘It would be better to do a than to do b’” (Davidson 2001, p. 39). Thus, an unconditional judgment expresses no more than what ought to be done without mentioning grounds for why the action ought to be done.
Davidson’s solution to the problem of weakness of will consists in showing that an agent who acts incontinently combines a special case of a conditional value judgment where “all things considered, x is better than y” with the unconditional judgment that “y is better than x.” P2 expresses an unconditional judgment. By P1, it is this unconditional judgment on which a person acts. D only mentions a prima facie judgment. Davidson’s account shows us that a conditional judgment cannot conflict with an unconditional one. Thus, the akrates makes two compatible judgments. Doing y is intentional because it is done from a reason provided by the unconditional judgment. At first glance, the unconditional nature of P2 went unnoticed; however, attention to the analogy with probabilistic reasoning has shown that this small verbal difference bears enormous theoretical weight.
Criticism of Davidson's Solution and a Possible Response
The question remains whether Davidson actually shows that weakness of will is possible. Several critics believe he has failed in this endeavor. For instance in “Davidson on ‘Weakness of the Will’” (1985), Paul Grice and Judith Baker argue that there is no way to create ‘all things considered’ judgments that is both consistent with conditional judgments generally and yields a satisfactory analysis of the problem of weakness of will. For Grice and Baker, it is implausible that someone should fail to reason from the ‘all things considered’ judgments to an unconditional judgment given that Davidson’s notion of ‘all things considered’ judgments is acceptable. In other words, Grice and Baker assume that ‘all things considered’ judgments are identical to ‘unconditional’ judgments. Hugh McCann (1998) goes a step further by arguing that if out of weakness of will an agent were to decide y, the agent would not consider to have judged doing so to be better than doing x, all things considered. That an agent considers y superior to the available alternatives does not seem plausible under the circumstances since the agent judged it would be better to do x. McCann also does not know what the “provenance of such an [all things considered] judgment would be” (1998, p. 225). To do the opposite of what an agent judges better than some other alternative would be an act not of irrationality, but of “sheer lunacy” (McCann 1998, p. 225). These criticisms entail that Davidson’s account is implausible. The main difficulty seems to be Davidson’s understanding of what an unconditional judgment is if it is not an all things considered judgment.
Those who believe Davidson’s account is implausible never seem to explain what plausibility is (Rescher 1974). The plausibility of a proposition captures the extent to which a proposition, if accepted as true, would fit within the wider setting of what we incline to accept. What is at stake is the acceptability or the unacceptability of a proposition, not how probable something is relative to what we know to be true. For example, when I know there are 2 pencils in my pocket, the probability that there are 3 pencils is very low; but, the plausibility of this prospect is not insubstantial. For there to be 100 pencils in my pocket is very implausible indeed. Plausibility of a proposition is a matter of depth of the revisions in the body of our knowledge that would be called for if that proposition were to turn out true. The more amenable a proposition is relative to the wider epistemic community determines whether or not one ought to accept a single proposition. On this account, Grice and Baker as well as McCann possess a seemingly unorthodox notion of plausibility. They would have us believe that plausibility is a matter of relation between propositions, not of the proposition simpliciter. As I have tried to show, plausibility involves singular propositions, whereby the more plausible a statement is means the more deeply we are committed to accepting that statement. So, critics who charge Davidson with an implausible account must give reasons why one of the principles is implausible than to offer a reason why the relation between principles is implausible.
Beyond the plausibility issue, some believe that his account is unsound. Such a critique would conclude that the following two statements are blatant contradictions:
A1 All things considered, it would be better to do y than to do x.
A2 It would be better to do x than to do y.
Davidson would say that the akratic agent judges that A1 and A2 are true while opponents point to the logical contradiction that the conjunction of A1 and A2 entails. Davidson would say that there is no contradiction (logical or otherwise) in A1 and A2. A1, on the one hand, is an ‘all things considered’ judgment, and A2, on the other hand, is a sans phrase judgment. Even if we accept Davidson’s explanation that A1 and A2 are not contradictories, the contradiction reappears in another case.
A contradiction arises if we tweak P2 such that:
P2´ If an agent judges that all things considered it would be better to do x than to do y, then he wants to do x more than he wants to do y.
The combination of P1 and P2´ surely contradict P3. It would logically follow from P1 and P2´ that an agent perform the action judged best on the basis of all available reasons. This renders P3 false. P2´ suggests that Davidson’s solution is as suspicious as it looks at first glance.
To support such a criticism one need only add a further principle:
P4 If an agent judges that all things considered it would be better to do x than to do y, then he wants to do x more than he wants to do y, if he wants to do either x or y.
P4 seems reasonable in light of P1, P2, and that there are akratic acts fitting Davidson’s definition. Unlike P1 and P2, P4 includes all statements that the agent judges relative to reasons, i.e. implicit or explicit reasons and some or all things considered reasons. The analogy with probabilistic reasoning should make this clear. For instance, if someone were to say, “Probably it will rain,” then the relevant considerations, e.g. “that the barometer is falling,” “that those rain clouds look threatening,” or “that the weatherman said so,” are unexpressed. Even though the relevant considerations are implicit, they are surely understood. In other words, it will probably rain all things considered; the critic would conclude that there is no judgment “Probably it will rain” sans phrase. According to the critic, a judgment deemed ‘better’ sometimes leaves considerations unexpressed. Davidson is wrong to think that these judgments are not always relative to some or to all relevant considerations. Therefore, Davidson’s conclusion that akratic acts are possible is unsound.
Davidson would find it difficult to accept P2´ for the simple reason that it is superfluous and unnecessary. P1 and P2´ are both all things considered judgments. The combination of P1 and P2´ contradict P3, but they are merely claims one may assert, not an argument. The reformulation of P2 does suggest that Davidson’s solution is suspicious because it has misconstrued his original position. The addition of P4 seems equally mysterious and forthrightly irrational. It is irrational to assume that an agent must necessarily judge all statements relative to reasons. If P4 were true, it seems an agent could perform only those actions that are the result of all the relevant reasons. An agent that judges all statements relative to reasons would have to account for reasons to do x, the reasons for reasons to do x, the reasons for reasons for reasons to do x ad infinitum. That a reason justifies an action must also be explained. Reasons for reasons to act should account for reasons to act. All things considered judgments should be able to account for these reasons if they are truly reasons that account for all information. Moreover, that the reasons are unexpressed, as this critic supposes, seems to be what judgments sans phrase are. Not only are judgments sans phraseunexpressed, they are unexpressible. The agent acts on these judgments – period. No more judgments are necessary. For an agent to judge that “x is best, all things considered” is to say that an agent judges x is the appropriate action considering all of the relevant alternatives. Judgments sans phrase, however, do not consider the alternatives. They are judged as if they are the only available options. So, this seems to somewhat vindicate Davidson of advancing an unsound argument (though other arguments may be lurking in the background).
On the possibility of the weakness of will
Confusion arises from whether one can hold that “x is better than y, sans phrase” while also holding that “y is better than x, all things considered.” To my mind, such a problem is the result of accepting the account of weakness of will simpliciter versus accepting a modal account of weakness of will. A modal account wants to express that something is possible. To express something is possible means to give an account that may occur in such-and-such a way. Davidson’s version seems to accord with the latter of the two options.
Since the contrast between all things considered judgments and judgments sans phrase generates such confusion, Davidson thinks it best to modify D, the original definition of weakness of will:
D´ An action, x, is incontinent provided that the agent has a better reason for doing something else: an agent does x for a reason r, but the agent has a reason r´ that includes r and more, on the basis of which the agent judges some alternative y to be better than x. (Davidson 2001, p. 40)
In revising the original definition, Davidson has eliminated the phrase all things considered. Having any amount of reasons for or against performing some action does not necessitate one to perform or not to perform the action. At the time of action, an agent may hold various reasons for doing x, but the agent may still do y. By doing y, the agent may or may not view doing y as the optimal choice; for the agent may still see x as the better of the two available options. The agent need only perform those actions deemed possible, which the agent cannot compare with other actions. An action is continent if that action’s performance is done for a reason, and there is no reason on which the agent judges some alternative action better. Davidson later says, “If r is someone’s reason for holding that p, then his holding that r must be, I think, a cause of his holding that p. But, and this is what is crucial here, his holding that r may cause his holding that p without r being his reason; indeed, the agent may even think that r is a reason to reject p” (Davidson 2001, p. 41). The first sentence seems fairly straightforward: the agent’s reasons for holding something cause the agent to hold it. The critical piece of Davidson’s reasoning comes next. When he says that an agent’s holding that r may cause the agent to hold that p, Davidson seems to say that r may possibly cause an agent to hold that p but the agent need not necessarily hold r to cause the agent to hold that p. For example, “wanting to talk about the Patriots game” may be a reason for Tom to “go to lunch with colleagues,” but Tom may just “go to lunch with colleagues” without “wanting to talk about the Patriots game.” In this sense, r may or may not be a reason for or against holding that p. Therefore, an agent may have a reason, r0, for an action, x, and a reason r´ that includes the set of r0, r1, r2, …rn for x, which allows the agent to judge some alternative y to be better than x, but y cannot be compared to other alternative actions, i.e. z, etc., because the reason, q, for doing y is exempt from comparing it to other reasons.
Weakness of will, then, is possible, even if P1 and P2 are true. Even if P1-P3 are incompatible, such incompatibility does not necessarily rule out incontinent actions. We can only conclude that incontinent actions are possible because of the difference between conditional and unconditional judgments as Davidson has so aptly pointed out. Incontinent actions may be deemed strange, unusual, or the like, but deeming them as such does not necessarily make them impossible.