A Counterexample to Setiya

In his book, Reasons Without Rationalism, Keiran Setiya posits the following necessary truth about intentional action:

Belief: When someone is acting intentionally, there must be something he is doing intentionally, not merely trying to do, in the belief that he is doing it.

Setiya’s requirement builds on Anscombe’s insight that “intentional actions are ones to which a certain sense of the question ‘why?’ has application.”(Anscombe [2000], Intentions. p. 11, §6.) In specifying precisely what that sense is, Anscombe notes that “this question is refused application by the answer: ‘I was not aware I was doing that’.” (Ibid.) Setiya takes this to suggest a conception of intentional action according to which an agent must know that she is performing a certain action in order to count as performing that action intentionally. Thus, we arrive at what may be called the Strong Knowledge Requirement for intentional action:

Strong Knowledge Requirement: For all agents, φ, if an agent is φ-ing intentionally then that agent knows she is φ-ing.

However, after considering Davidson’s example of the teacher who is intentionally making ten carbon copies as he writes even though he is unsure that he is pressing hard enough to successfully do so, Setiya concludes that the Strong Knowledge Requirement is unsound. Since there are times we do not know that we are successfully performing an action we are intentionally performing, such knowledge cannot be a necessary condition for intentional action.

The first revision Setiya makes to Anscombe is to switch from a knowledge to a belief requirement. This yields what may be called the Strong Belief Requirement for intentional action:

Strong Belief Requirement: For all agents, φ, if an agent is φ-ing intentionally then that agent believes she is φ-ing.

However, as Setiya acknowledges, the Strong Belief Requirement does little better than the Strong Knowledge Requirement vis-a-vis the carbon copy case. Insofar as the carbon-copier is unsure about whether or not he is making ten copies, he does not believe that he is making ten copies. Setiya therefore proposes a second modification to the Strong Knowledge Requirement. He claims that to count as φ-ing intentionally, one need not believe that one is φ-ing. One only needs to believe that one is performing some action ψ, where ψ is either identical with φ or an intentional action one is performing with the end of φ-ing. Moreover, he holds that one’s belief that one is ψ-ing must be true; to wit, ψ must be an action one is actually performing rather than merely attempting to perform. Thus, we arrive at what I take to be Setiya’s considered position with respect to intentional actions:

Moderate Belief Requirement: For all agents, φ, if an agent is φ-ing intentionally then that agent believes truly that she is ψ-ing, where ψ-ing is either identical to φ-ing or an intentional action performed with the end of φ-ing.

It should be clear that the Moderate Belie Requirement is nothing but a restatement Belief. This revision of Anscombe's (alleged) Strong Knowledge Requirement allows Setiya to successful address the carbon copy case since there is an intentional action (i.e., pressing hard while writing) that the carbon-copier believes truly that he is performing, and which is performed with the end of producing ten copies.

I wish to argue that Belief fails to capture a necessary truth about intentional action. To this end, I will be attempting to construct a counterexample to Belief; a case in which we would plausibly regard an agent as acting intentionally even though the agent fails to meet the necessary condition it specifies.

The Prosthetic Limb Example*:
Consider the case of an arm-amputee, Jesse, who has a thought-controlled prosthetic arm grafted to his shoulders. Let us suppose that Jesse has to demonstrate the functionality of his thought-controlled arm to a Research and Development panel. However, just before the demonstration, Jesse gets an anonymous letter saying that the researchers, out of fear that their funding will be cut, have conspired to trick him into thinking that his prosthetic arm is functioning properly even though it is not. According to the anonymous letter, the researchers will ask Jesse to perform a number of tasks and observe him closely for an indication that he is about to perform the requested action. Then they will remotely cause the arm to perform the various tasks using a wireless signal from a computer. Thus, according to the anonymous letter, while it would appear to him that he is controlling his prosthetic arm with his thoughts, it will actually be the researcher’s computers that will be determining the arm’s movements.

However, let us suppose that (unknown to Jesse) the anonymous letter is completely unreliable, and that the researchers have concocted no such plot. All the movements his prosthetic arm makes are in fact being caused by his thoughts rather than by the scientist. As he is standing before the panel, a ball is thrown towards Jesse and he catches it with his prosthetic arm. He is then asked to throw the ball, and he complies. Moreover, let us assume that the thought process preceding the movement of the thought-controlled arm are of the same kind as that which would precede the movements of Jesse’s normal (i.e., non-prosthetic) arm. However, since he is unsure about the reliability of the anonymous letter, Jesse remains unsure that it is his thoughts that are causing the arm to catch the ball. Even as he is performing the action he can’t help but wonder if it is actually the researchers who are controlling the arm’s movements with a remote device. In short, there is no action (i.e., moving his arm, catching the ball, throwing the ball) that Jesse believes he is performing. Still, it seems perfectly natural to say that Jesse caught and threw the ball and that he did so intentionally. The upshot is that, contra Belief, Jesse intentionally catches and throws the ball even though there is no action that he believes he is performing.

There are two strategies for resisting the unpropitious consequences of the Prosthetic Limb Example that appear worth considering:

(Strategy 1): Argue that Jesse has not performed an intentional action when he catches or throws the ball.

(Strategy 2): Grant that Jesse has performed an intentional action, but argue that there is an intentional action that he believes he is performing with the end of catching and throwing the ball.

I believe that (Strategy 1) is moribund. Firstly, we may say that Jesse either intended to catch and throw the ball, or he did not so intend. These exhaust all the relevant possibilities. Now, it seems highly implausible to say that Jesse did not intend to catch the ball. Clearly, his catching and throwing the ball was no accident. Nor was it the side-effect of some other action Jesse was performing. Moreover, we may safely assume that Jesse went before the panel with the intention of performing the various tasks asked of him (even if he was unsure he would be the one performing them). Additionally, when the ball was thrown towards him, it is clear that Jesse meant to catch it; this was the goal he had in mind when he moved his prosthetic arm to intercept it. Once it has been acknowledged that Jesse intended to catch and throw the ball, it is natural to regard his intention as the cause (in the broadest sense of the term) of his arm’s movements. In fact, since (ex hypothesi) the arm’s movements are not being controlled by the researchers, it is unclear what other explanation there could be of his catching and throwing the ball besides Jesse’s intention to catch and throw it.

Secondly, suppose it is later revealed to Jesse that the anonymous letter was unreliable and that he was actually controlling the movements of the prosthetic arm. In such a case, I believe we can imagine Jesse thinking to himself, “so I did catch the ball after all!” In other words, it is plausible that Jesse would take ownership of the action in the way one would take ownership of something one intentionally performed. Moreover, it would be implausible to suggest that Jesse’s belief that he performed the action retroactively made his action intentional. Jesse’s belief does not change the (metaphysical) status of his action from unintentional to intentional; at most, it changes his knowledge of the action’s status. Thus, we must conclude that Jesse’s actions were intentional all along, if we are to hold that it is intentional at all. It follows that Jesse’s actions were intentional even when he did not believe he was performing them.

According to (Strategy 2), there is in fact some intentional action that Jesse believes he is performing—namely, whatever thoughts caused the movement of his thought-controlled prosthetic arm. Unfortunately, (Strategy 2) also seems moribund. Ex hypothesi, Jesse’s prosthetic arm is controlled by the same kind of thought processes that are at play when he moves his normal arm. There is some debate over whether thoughts—particularly of the kind that features in the aetiology of bodily movements—may be considered intentional actions. But let us grant, if only for the sake of argument, that the relevant thoughts are themselves a kind of intentional action. Even so, the present proposal only seems remotely plausible if we identify the thought in question with ‘trying’ to catch or throw the ball, in the sense of ‘trying’ that accompanies all cases of intentional action. This is the sense of ‘trying’ that Davidson attributes to the carbon-copier, and which Davidson takes to be sufficient for the carbon-copier’s actions to be intentional. Thus, if we buy into Davidson’s framework, we can easily accommodate the intuition that Jesse’s actions are intentional.

However, this is not an option available to Setiya, who explicitly denies that ‘trying’ (in the above sense) fulfils the criterion imposed by Belief. He writes:

Despite what Davidson suggests, it is not enough that the carbon-copier is intentionally trying to make ten copies, in the paradigm sense of “intentional action” that involves belief. He is and must be doing specific things—for instance, pressing hard on the paper—in that paradigm sense.

In brief, Jesse’s act of trying to catch the ball (even if it is regarded as an intentional action) fails to meet Setiya’s specifications. Consequently, Belief entails that Jesse does not catch the ball intentionally. It follows that, by Setiya’s own lights, (Strategy 2) gets things wrong. Assuming that (Strategy 1) and (Strategy 2) exhausts all remotely plausible strategies for responding to the Prosthetic Limb Example, it remains a counterexample to Belief.

Admittedly, the Prosthetic Limb Example is an unusual case. Moreover, it is clear that in the vast majority of cases of intentional action, the agent is not in the position that Jesse finds himself in. Consequently, I do not see the Prosthetic Limb Example as posing a challenge to the claim that when we act intentionally we prototypically know that we are acting. Moreover, since I take Anscombe to be offering a prototypical generalisation, I do not believe the Prosthetic Limb Example represents a refutation of Anscombe’s account of intentional action. However, what Setiya purports to provide is not a prototypical generalisation, but a necessary truth. Unlike prototypical generalisations, necessary truths allow for no exceptions. Thus, as unusual as the Prosthetic Limb Example may be, it is sufficient to undermine Setiya’s claim that Belief is a necessary truth.

*This example is loosely based on the real life case of Jesse Sullivan. Here is a short video of Jesse's arm in action:

Watch CBS Videos Online

Might Intentions be Reasons? (Ben Mitchell-Yellin)

In this post, I want to gesture at a sense in which intentions might be reasons that is consistent with the view defended by John Brunero (“Are Intentions Reasons?” Pacific Philosophical Quarterly 88: 424-44), that intentions do not provide reasons for actions or for adopting other intentions. I will begin by briefly presenting Brunero’s view. Then I will make a distinction between different “directions” in which a reason may “push” and suggest that intentions might be reasons that push in one direction even if not the others.

Brunero rejects both the intentions-provide-reasons view and the tie-breaker view, according to which intentions are reasons only in tie-break situations. He puts forward the following positive view:

[I]n cases where we are dealing with some already reasonable end, there is a reason to do what would facilitate that end (a reason that exists not because you intend the end, but because the end is reasonable), but your intending the end may be relevant to whether this reason transfers to the specific actions which are necessary, but not sufficient, for the realization of the end.

Brunero’s idea of a reason transferring from an adopted end to a necessary (but not sufficient) means relies on the notions of the “facilitative principle” and “facilitating plans,” borrowed from Raz. A facilitating plan is a set of actions that are individually necessary and jointly sufficient for achieving some end. According to the facilitative principle, given reasonable end E, agent A has a reason to perform each individually necessary (but not sufficient) means M1, M2, …, Mn, as part of his facilitating plan P. But this does not mean that A has a reason to perform M1 by itself. That “would seem a pointless and unreasonable waste of time.” This is where intentions affect our reasons. A’s intention to E “provides us with some assurance that he will undertake the other parts of the plan that, along with [M1], will jointly suffice to bring it about that he can” E. So A’s adoption of the intention to E transfers the reason for P as a whole to each of its individual parts M1, M2, …, Mn.

Notably, on Brunero’s view, the intention to E does not give A any new reasons. Instead, it affects the structure of the reasons he has. Both E and E’ may be reasonable ends. So A has reason to undertake both facilitating plans P and P’. But supposing A adopts E, and not E’, we do not want to say that he has a reason to M and M’, where these are individual parts of P and P’, respectively. Brunero’s view allows us to say that A’s adopting E, and not E’, makes it the case that A has a reason to M (as a part of P) but not M’, and this is so because A’s adopting E makes it reasonable to suppose that A will complete all the necessary and jointly sufficient means to E, of which M is one. I think this view is reasonable. I also think it allows that intentions might be reasons, albeit in a different sense than concerns Brunero.

Brunero follows the literature in concerning himself with the question whether intentions might be considerations that favor performing actions (that are necessary (but not sufficient) means to executing the intention) or adopting other intentions. We might say that in the first case the question is whether intentions are reasons that “push down” and in the second case whether intentions are reasons that “push across.” Such talk makes sense if we think of levels of rational agency (perhaps along the lines of Bratman’s theory of planning agency). We might delineate the various levels as follows. At bottom, we have facts. Then we have actions, facilitating plans, intentions, non-facilitating plans and policies, in ascending order of hierarchy. One natural question, given this hierarchy, is whether the adoption of an intention exerts rational pressure that “pushes down” to the level of facilitating plans. Does A’s adopting the intention to E give her a reason to M, where M is a necessary (but not sufficient) means to E? This is the sort of question that concerns Brunero. He gives a negative answer. A’s adopting E does not give her a reason in this sense. She already has a reason to M, but only as a part of facilitating plan P. A’s adopting E transfers the reason for P to each of the jointly sufficient actions that constitute it. One may even think this claim can be generalized, such that no reasons “push down.”

But, it seems to me, even if the general claim is correct in answer to the above question, we might also ask a different question. We might wonder whether the adoption of an intention to E exerts rational pressure that “pushes up” to the level of non-facilitating plans. Does A’s adopting the intention to E give her a reason to adopt non-facilitating plan X, where E is a necessary (but not sufficient) means to X? Perhaps we can give a positive answer. So intentions can be reasons in a sense. They can be considerations that favor adopting non-facilitating plans. The claim that reasons “push up” is not at all foreign. Facts are commonly thought to favor actions, for example. So a negative answer to the general question whether reasons ever “push up” is not likely forthcoming. This adds interest to the question about intentions in particular.

Let us take as basic the notion of a reason as a consideration that favors something (following Scanlon). If A’s adopting an intention favors something (in a way that thing was not favored before the adoption of the intention), then we can say that A’s adopting the intention gives A a reason. This is consistent with Brunero’s view because he accepts the claim that reasons are basic and denies that A’s adopting an intention favors anything (in a way that thing was not favored before the adoption of the intention). I want to suggest (but not argue for the claim) that once we notice the above hierarchy, we might see that the adoption of an intention might favor something (in a way that thing was not favored before), namely, a non-facilitating plan. We are familiar with the notion that a fact can favor an action. So we are familiar with the notion of reasons that “push up” from facts to actions. Indeed, it is because some fact F already favors M that we need not take A’s adopting E to favor M in order to say that A has a reason to M. And we can say that E is reasonable because it is favored by certain facts. This is another familiar way that reasons “push up.” So what stands in the way of our saying that another way reasons “push up” is by intentions favoring non-facilitating plans?

Here is one possible objection. We might think that only facts can stand in the favoring-relation to things (e.g., actions, intentions, etc.). The two familiar ways that reasons “push up” both appeal to facts favoring something. We might say that facts, not intentions, favor non-facilitating plans in just the same way that facts favor actions and intentions. Call this a “buck-passing” account of rational agency. At each level, the relevant thing is reasonable only if favored, but it is always facts that do the favoring. So intentions do not favor non-facilitating plans, facts do. Thus, intentions do not give reasons.

This is an interesting objection. Let me conclude by suggesting one reason why we might not want a buck-passing account of rational agency. I think we want our adopted goals to factor into the reasonableness of our higher-order agential attitudes. We want our intentions to affect the reasonableness of our non-facilitating plans and our non-facilitating plans to affect the reasonableness of our policies. This is not to deny that the facts affect the reasonableness of our non-facilitating plans and policies. But the facts may not (do not?) determine uniquely reasonable non-facilitating plans and policies. And this is where we might find that intentions are reasons. That A adopts an intention to E might favor her adopting certain reasonable (according to the facts) non-facilitating plans over others.

Why Care About "God"?

The 93rd Philosopher's Carnival

is here.

Setiya On Intentional Action

In this blog post I wish to articulate what I take to be the primary objection to Kieran Setiya's account of intentional action, as described in the first half of his book, Reasons without Rationalism. I begin by adumbrating a few of my own commitments, followed by a summary of Setiya's position, and I conclude with critical remarks.

I hold that intentional action is prototypically goal-directed action. On this view, the type of goal-oriented behaviour a cat engages in when it stalks a bird counts as intentional. By contrast, purely reactive behaviour, such as when a cat reflexively withdraws its paws away from a sharp object, is non-intentional. In sum, I take it to be paradigmatic of intentional action that it is purposive rather than merely reactive.

My primary reason for adopting the present definition of ‘intentional action’—one that conceives of it in terms of goal-directed action—is my belief that it closely matches our quotidian conception. Ordinarily, we identify an agent’s intention with the aim, purpose or goal they have in mind when carrying out some action. Moreover, I maintain that an agent may perform an action with a certain goal in mind even if that agent does not (or cannot) conceive of that goal as such. To conceive of a goal as such requires that one possess and deploy the concept of a goal. Thus, an account of intentional action that requires that an agent conceive of their goal as such would preclude non-linguistic animals—that lack such concepts—from acting intentionally. By contrast, to have a goal in mind is to be aware of one’s goal in the same sense in which one may be aware of the content of one’s perceptual experience. Since non-linguistic animals may be aware of the content of their perceptual experiences, there is nothing in the present account of ‘having a goal in mind’ that precludes its application to non-linguistic animals.

My second reason for defining ‘intentional action’ in the way that I have is that it roughly corresponds with that of Anscombe in her landmark text, Intention. Admittedly, Anscombe is not committed to the claim that intentional action just is goal-directed action since she holds that an agent may act intentionally even though she has no goal or end in view. However, she does take goal-directed action as the paradigm case of intentional action, such that there would be no such thing as intentional action if we did not sometimes act with an end in view. This is a subtlety in Anscombe’s account that I cannot fully explore here. But it is sufficient for our present purpose to note that Anscombe and I agree with respect to there being a conceptual connection between intentional action and goal-directed action. Significantly, Anscombe and I both concur that one may correctly ascribe intentions to non-linguistic animals. She puts the point as follows:

Since I have defined intentional action in terms of language—the special question ‘Why?’—it may seem surprising that I should introduce intention-dependent concepts with special reference to their application to animals, which have no language. Still, we certainly ascribe intention to animals. The reason is precisely that we describe what they do in a manner perfectly characteristic of the use of intention concepts. . . . the cat is stalking the bird in crouching and slinking along with its eye fixed on the bird and its whiskers twitching. . . . Just as we naturally say ‘The cat thinks there is a mouse coming’, so we also naturally ask: Why is the cat crouching and slinking like that? and give the answer: It’s stalking that bird; see, its eye is fixed on it. We do this, though the cat can utter no thoughts, and cannot give expression to any knowledge of its own action, or to any intention either.

Since Anscombe is the locus classicus of the contemporary discussion of intention, I take her usage of the term to have the greatest claim to philosophical orthodoxy. Of course, we may find the need to make adjustments to her conception along the way; but I think one can hardly go wrong (from a methodological point of view) in taking her as a starting point. Moreover, I will take Anscombe’s observation that we ordinarily ascribe intentions to non-linguistic animals as a touchstone for determining whether or not a particular theorist is working with the philosophically orthodox conception of ‘intentional action’. My reasons for this are far from arbitrary. It rests on the thesis that those theorists who deny the ascription of intentions to non-linguistic animals are actually working with a very different concept (and are therefore talking about something quite different) to those who affirm such ascriptions. When this fact is combined with a certain lack of self-awareness with regards to the differences in the concepts being deployed, the upshot is that theorists on both sides are often simply talking past each other. The preceding claim is not one I can fully defend here; so a dogmatic statement of my position will have to suffice. However, the central criticism I will be advancing against Setiya in this post does not depend on the undefended assumption.

The conception of intentional action employed in this blog post—one that defines intentional action in terms of goal-directed action—differs from that of Setiya, who sees the two terms as picking out fundamentally different domains of agential activity. According to Setiya, an agent φs intentionally only if that agent has the higher-order desire-like belief that she is φing for a reason. Setiya’s requirement seems to build on Anscombe’s insight that “intentional actions are ones to which a certain sense of the question ‘why?’ has application.” In specifying precisely what that sense is, Anscombe notes that “this question is refused application by the answer: ‘I was not aware I was doing that’.” Setiya takes this to suggest a conception of intentional action according to which an agent must know that she is performing a certain action in order to count as performing that action intentionally. (I will refer to this as the knowledge requirement for intentional action.) However, after considering Davidson’s example of the teacher who intends to make 10 copies, but is unsure if he is pressing hard enough to successfully do so, Setiya concludes that the knowledge requirement is too strong. The assumption seems to be that since there are times we do not know that we are successfully performing an action we are intentionally performing, such knowledge cannot be a necessary condition for intentional action. He therefore falls back to a belief requirement, according to which one acts intentionally only if one believes one is performing said action.

Setiya’s makes a second modification to Anscombe’s knowledge requirement. He claims that one not only believes that one is performing a certain action, but that one also believes that one is performing it for some specific consideration that one takes as one’s reason (insofar as one is performing it for a reason at all). This leads him to distinguish between two aspects of what it means to take the consideration that P as one’s reason to φ. The “practical” aspect involves a desire-like attitude towards the proposition: the consideration that P is my reason to φ. The attitude in question is “desire-like” in that it motivates one to φ. The “epistemic” aspect of taking the consideration that P as one’s reason is the non-observation based belief that the consideration that P is my reason to φ. The practical and epistemic aspects combine to form what Setiya refers to as a “desire-like belief”:

Taking something as my reason is a kind of “desire-like belief”. It is a belief-like representation of P as my reason to act, and at the same time a decision to act on that reason, something by which I am led to do so.

In order to avoid the charge of circularity, Setiya later modifies the content of the desire-like belief from the proposition: the consideration that P is my reason to φ, to the proposition: my belief that P is my reason to φ. In short, the desire-like belief is a second-order belief; it takes as its object the belief that P is one’s reason to φ.

In an astonishing display of pluck, Setiya describes his approach as a “simple psychological theory” (Italics mine), according to which “taking something as one’s reason is a matter of taking one’s belief in that reason to play a causal-motivational role in explaining one’s action”. Whatever the merits of such an account of intentional action, it is clear that it precludes the possibility that non-linguistic animals may act intentionally. The requirement that one believe that one’s belief that P is playing a “causal-motivational role in explaining one’s action” is not one that non-linguistic animals (and even a few philosophers) can meet. Moreover, Setiya’s account precludes the possibility that intentional action is prototypically goal-directed action since a wide range of goal-directed actions (e.g., those performed by non-linguistic animals and pre-linguistic human infants) involve no such second-order beliefs. This clearly puts Setiya at odds with the account of intentional action we find in Anscombe, and (therefore) out of step with what I have been calling the philosophically orthodox conception. I believe this represents a problem for Setiya’s account because when he criticises Anscombe and advocates of the belief-desire model for their accounts of intentional action, it is not clear that he is talking about the same thing as those he criticises.

More importantly, the fact that Setiya’s account of intentional action (1) denies the prototypical connection between intentional action and goal directed action, and (2) precludes the ascription of intentions to non-linguistic animals, suggests that his account is revisionary with respect to our quotidian conception. This presents Setiya with the following dilemma. On the one hand, if his account of intentional action is supposed to coincide with our ordinary usage of the term, then his theory implies that the vast majority of competent English speakers are simply mistaken when they ascribe intentions to non-linguistic animals. This may of course be the case, but we would need very compelling reasons for thinking that this is so along with an error theory of some kind. Setiya provides neither. On the other hand, if his account is not supposed to coincide with our ordinary usage of the term, then the relevance of his account is called into question. Why should we be concerned with this new concept Setiya is attempting to introduce? Moreover, why does he risk confusing the reader by using the expression ‘intentional explanation’ to refer to this novel conception, without the least indication that he is using it in an unconventional way?