Friday, 27 June 2008

Davidson on Weakness of Will (Part 3)

In my last two posts I described Davidson as advocating a two-stage model of practical reasoning. In this post, I wish to limn an alternative, and what I believe to be a more charitable, interpretation of Davidson’s account. The central insight driving this alternative reading of Davidson is the intuition (originally due to Elizabeth Anscombe) that there is a conceptual connection between intentional action and the applicability of the “why?” question to one’s actions. (I owe this point to Agnes Callard.) To ask “Why did Sue wave at Pam?”, is to ask for Sue’s reason for acting as she did. This question only makes sense if Sue waved her hand deliberately; that is, if her actions were intentional. Hence, for an agent to act intentionally is for that agent to act for a reason.

Davidson defines a reason to φ as the combination of a desire (or pro-attitude) for actions of a certain kind and a belief that φ is an action of that kind. For example, if Sue’s desire to signal her friend Pam in a crowded pub motivates her to wave her hand, then her reason for waving consists in her desire to signal her friend and her belief that waving would be a signalling of her friend. On this belief-desire model, an agent’s (primary) reason for performing an action may be restated in a practical syllogism comprised of a major premise (M), a minor premise (m) and a conclusion (C):
(M): ψ (signalling Pam) is desirable
(m): φ (waving) is an instance of ψ
(C): Therefore, φ is desirable
Thus, for each action one can construct a practical syllogism showing that the action was desirable from agent’s perspective. According to Davidson, what it means for an action to be desirable is that the agent would perform this action, provided that there are no overriding desires:
If my thesis is correct, someone who says honestly ‘it is desirable that I stop smoking’ has some pro attitude towards his stopping smoking. He feels some inclination to do it; in fact he will do it if nothing stands in the way, he knows how, and he has no contrary values or desires. (Davidson in "Intending")
I will refer to the practical syllogism that corresponds to the reason for a certain action as its rationalisation. Two important points are worth noting with respect to rationalisations. First, rationalisations are explanations of an agent’s actions that reflect the point of view of the agent. Second, while a rationalisation spells out the reasons for an agent’s actions, it does not specify whether or not the agent’s reasons are good reasons. Thus, Davidson writes:
When we talk of reasons in this way, we do not require that the reasons be good ones. We learn something about a man’s reasons for starting a war when we learn that he did it with the intention of ending all wars, even if we know that his belief that starting a war would end all wars was false. Similarly, a desire to humiliate an acquaintance may be someone’s reason for cutting him at a party though an observer might, in a more normative vein, think that was no reason. The falsity of a belief, or the patent wrongness of a value or desire, does not disqualify the belief or desire from providing an explanatory reason. (Davidson in "Intending")
Generally, rationalisations enable us to see the actions of an agent as reasonable in the light of an agent’s beliefs and desires. But that is not to say that the agent’s actions would be reasonable if all her beliefs and desires were considered. In the cases of weakness of will, an agent’s actions are only reasonable in the light of specific beliefs and desires, but not in the light of all her beliefs and desires. (I will have more to say about this below.)

In light of the above considerations, I propose that we see Davidson as offering a two-level (rather than two-stage) model of practical reasoning. On this reading, Davidson’s account is seen as providing two levels of explanation for intentional actions: a rational explanation and a causal explanation. The rational explanation of an agent’s actions consists in a practical syllogism that rationalises the action. The reasons an agent has for acting must, if they are to rationalise the action, be the reasons on which she acted. That is, the reasons must have played a causal role in the aetiology of the action. This causal explanation, in which an agent’s reason is seen as the cause of her action, constitutes the second level of explanation.

To rationalise an action is to show how the action is judged to be best, from a certain perspective of the agent. As such, a rationalisation is to be identified with a prima facie judgement which is relative to the set of reasons from which the agent ultimately acts. However, the prima facie judgement from which an agent acts (that which rationalises her actions) may not be the only prima facie judgement made by the agent. For example, in cases of moral conflict, an agent has reasons for two (or more) competing courses of action. On such occasions, the agent makes more than one prima facie judgement, each one relative to the reasons the agent has for each competing course of action. However, since there can only be one primary reason for any given action—namely, the reason which is the cause of the action—then, in cases of moral conflict, only one of the prima facie judgements can correspond with an agent’s primary reason. Combining this claim with our earlier observation that each prima facie judgement is relative to some set of reasons, we may say that the prima facie judgement that is relative to an agent’s primary reason is the one that rationalises the agent’s actions.

On the two-level model, a prima facie judgement is expressive of an agent’s reasons in their evaluative capacity. However, as we already noted, the very reasons that factor into an agent’s evaluations may also cause (or motivate) the agent to act. In there motivational capacity, an agent’s reasons express her sans phrase judgements. Within the two-level explanatory framework, the very reason that features in an agent’s evaluations may also feature in her motivations, albeit in two different capacities.

With these considerations in place, let us return to the problem of continence and incontinence. Recall, both continence and incontinence only arise in cases of moral conflict; occasions on which an agent makes two (or more) prima facie judgements in favour of opposing courses of action. An agent counts as continent when her sans phrase judgement corresponds with the prima facie judgement that is relative to her total set of reasons. An agent counts as incontinent when her sans phrase judgement corresponds with a prima facie judgement that does not reflect her all-things-considered judgement. For example, consider the case of Mr. Smith who makes the following three prima facie judgements:
(c) Prima facie, having a third glass of wine is better than not having a third glass of wine, given r1.

(d) Prima facie, not having a third glass of wine is better than having a third glass of wine, given r2.

(e) Prima facie, not having a third glass of wine is better than having a third glass of wine, given R (where R represents his total set of reasons).
If Mr. Smith were to make the sans phrase judgement that, “not having a third glass of wine is best”, then his sans phrase judgement would correspond with his all-things-considered judgement, as represented in (e). In such a case, Mr. Smith would count as acting continently. However, if Mr. Smith were to make the sans phrase judgement that “having a third glass of wine is best”, then his sans phrase judgement would fail to correspond with his all-things-considered judgement. In such a case, Mr. Smith would count as acting incontinently. However, even if Mr. Smith were to make the latter sans phrase judgement, his actions would still correspond with one of his prima facie judgements, namely that represented in (c). Thus, it is still possible to give a rationalisation of Mr. Smith’s actions, albeit one that does not reflect his all-things-considered judgement. This is the upshot of our earlier observation that rationalising an action requires that we show that the action is based on reasons, not that is based on good reasons. As such, Mr. Smith (or the incontinent agent) still counts as acting intentionally.
With the two-level reading now on the table, we can return to the earlier worry about there being a motivational gap in Davidson’s account. Recall, the motivational gap arose because in order to preserve the consistency of (P1)-(P3), Davidson had to picture prima facie judgements (which are relative to an agent’s reasons) as purely evaluative, while the sans phrase judgements (which implicate the desires of agent) are seen as purely motivational. The upshot was that on the two-stage model an agent could not be motivated by her reasons which (on that picture) were relegated to her non-motivating prima facie judgements.

On the two-level model, the distinction between the purely evaluative prima facie judgements and purely motivational sans phrase judgements remains. However, the prima facie and sans phrase now represents two dimensions of explanation for a single piece of practical reasoning. On the two-level model, the same reasons that features (in their evaluative capacity) in the agent’s prima facie judgements may also feature (in their motivational capacity) in the agent’s sans phrase judgements. The upshot is that on the two level-reading, an agent’s reasons can be said to cause (or motivate) the agent to act, thus avoiding the motivational gap that threatened the two-stage model.

Friday, 13 June 2008

Davidson on Weakness of Will (Part 2)

In my previous post, I suggested that Davidson's bipartite account of practical reasoning is able to preserve the consistency of (P1)-(P3). However, there is a worry that Davidson’s solution leaves it unclear how the weak-willed agent can be said to be motivated by her reasons. This is the objection I wish to consider in the present post.

According to Davidson, the akrates acts for a reason, but not from her all-thing-considered judgement. Recall that in our example, Mr. Smith has three prima facie judgements—(c), (d) and (e)—each of which is relative to some set of reasons:
(c) Prima facie, having a third glass of wine is better than not having a third glass of wine, given r1.

(d) Prima facie, not having a third glass of wine is better than having a third glass of wine, given r2.

(e) Prima facie, not having a third glass of wine is better than having a third glass of wine, given R.
However, Mr. Smith (being weak-willed) ultimately decides to have a third glass of wine, against his all-things-considered judgement.

In the case of the continent agent, the action generating sans phrase judgement corresponds with the agent’s all-things-considered prima facie judgement. That is, the continent agent acts from her total set of reasons. However, in the case of Mr. Smith, his sans phrase judgement corresponds with (c) rather than (e). Hence, Mr. Smith (i.e., the incontinent agent) acts from a subset of his reasons, r1, instead of his total set of reasons, R. Herein, according to Davidson, lies the irrationality of the incontinent agent; she fails to perform the action judged best on the basis of all her reasons.

However, as limned above, there seems to remain a gap in Davidson’s account. On his picture, an agent’s prima facie judgement is always relative to some reason. However, an agent’s sans phrase judgement is not. This opens up a lacuna between one’s reasoned-based evaluations and one’s action generating judgements. Davidson attempts to fill this gap via what he refers to as the “principle of continence”:
(POC) Perform the action judged best on the basis of all available relevant reasons.
Davidson insists that (POC) is a principle that is binding on every rational agent, the violation of which constitutes the fundamental shortcoming of the akrates. Hence, Davidson believes that there is a normative imperative to always act from one’s total reasons. Strictly speaking, as worded by Davidson, (POC) is unable to bridge the logical gap between an agent’s prima facie judgements and her sans phrase judgements. However, we can equip it for this purpose by rephrasing it as follows:
(POC*) Do what is prima facie better, given R.
With the revised version of the principle of continence in place, the logical gap between an agent’s prima facie and sans phrase judgments has been bridged.

However, there still remains a motivational gap; a fact which the possibility of incontinence illustrates. If the principle of continence were sufficient to bridge the motivational gap between an agent’s prima facie and sans phrase judgements then any agent who was rationally committed to the principle of continence would be immune to weakness of will. However, such a conclusion seems implausible since it would imply that the akrates suffered from an extreme intellectual defect. Recall, Davidson invokes the principle of continence as a rule that every rational agent should accept as such. Consequently, to suggest that the akrates believes that she is not rationally obliged to act in accordance with the principle of continence is, by Davidson’s lights, to drum her out of the ranks of rational agents altogether

Davidson’s distinction between prima facie and sans phrase judgements drives a wedge between an agent’s reason-based practical assessments and desire-based motivations. On the one hand, one’s prima facie judgements are always relative to a set of reasons while one’s sans phrase judgements are not. On the other hand, only sans phrase judgements, according to Davidson, are expressive of one’s wants and therefore only sans phrase judgements can be said to motivate. The upshot seems to be that, on Davidson’s account of practical reasoning, one cannot be directly motivated by one’s reasons.

Another way to make this problem vivid is by posing the question: is (POC*) a prima facie or sans phrase judgement? There two considerations that tell against it being the former. First, it is not of the right logical form. (Although (POC*) represents a judgement about an agent’s prima facie judgements, it is not itself a prima facie judgement.) Second, if it were a prima facie judgement (assuming we were able to reformulate it somehow), it would be unable to bridge the motivational gap between Davidson's two stages of practical reasoning. (POC*) must implicate the desires of the agent if it is to serve the motivational role necessary. However, as we noted above, prima facie judgements fail to implicate the desires of an agent. On the other hand, (POC*) cannot be a sans phrase judgement since it does not consistently generate action (as the possibility of incontinence illustrates).

The suggestion that there is a motivational gap between an agent's prima facie and sans phrase judgements seems problematic since it seems to fly in the face of Davidson’s foursquare insistence that one is caused to act by one’s reasons.