Most proponents of the standard-causal approach acknowledge that the possibility of deviant causal chains constitutes a serious problem and that the plausibility and success of the view depends on whether that problem can be resolved convincingly.41 In this section I will first say more about the connection between the problem of causal deviance, acting for reasons, and agential control. Then I will introduce and discuss different forms of causal deviance, and, finally, I will propose a solution, which is based on guidance by and responsiveness to the contents of reason-states.
In the previous section we saw that the standard-causal theory can provide a viable account of an agent’s exercise of control in terms of the agent’s being guided by and responsive to reasons. It emerged that the problem of causal deviance is related to the issue of providing an account of acting for reasons and agential control. Examples such as Davidson’s climber show that the fact that an agent’s mental states and events cause and rationalise a bodily movement does not guarantee that the agent acts for reasons. Hence, the causal and rationalising relation is not sufficient for
40 This mistake is particularly salient in Velleman’s position. Velleman says that the problem with
Frankfurt’s hierarchical model of agency forces us to ‘look for mental events and states that are functionally identical to the agent’ (Velleman, 2000, p. 137).
41 Compare for instance Brand, 1984; Bishop, 1989; Davidson, 1980; Enç, 2003; Mele, 2003; Searle,
acting for reasons. The same holds for action and agential control, because the standard-causal model construes action and control in terms of acting for reasons. Examples of basic deviance show that the fact that a movement is caused and rationalised by reason-states is sufficient neither for action, nor for an agent’s exercising control. What is missing, as I suggested, is that the movement is guided by
and responsive to the agent’s reason-states. According to that suggestion, to amend
the account of acting for reasons and agential control in the way outlined is to provide a solution to the problem of basic deviance. That means that causation in the right—that is, non-deviant—way is causation that meets the conditions of guidance and reason-responsiveness.
Causal Deviance
It is common to distinguish between three different types of deviant causal chains in the theory of action. Davidson’s climber example, which has been introduced above, is a case of basic deviance, which is the most troublesome kind of causal deviance. Besides that we can distinguish between consequential deviance and second-agent
deviance.42
The former is a second main category of causal deviance—along with basic deviance—, whereas second-agent deviance can be treated as a special case of basic deviance. In order to distinguish these two further kinds, I need to say more about basic deviance.43
In all cases of deviance, some control-undermining state or event occurs between the agent’s reason-states and an event produced by that agent. What distinguishes cases of basic deviance is that the control-undermining event occurs between the reason-states and an agent-involving event that would constitute a basic action.44 Davidson’s climber example is a case of basic deviance. The climber rids himself of the weight and danger by loosening his hold on the rope. If the climber performed an action, the movement of loosening his hold on the rope would be a basic act. The state
42 I borrow the term ‘basic deviance’ from Bishop, 1989, p. 132, and the term ‘consequential deviance’
is taken from Brand, 1984, p. 23.
43 I may not do justice to all the subtleties of the phenomenon of deviance in action theory. It may be
possible to introduce some further sub-categories that require modified responses. The exposition, though, covers the standard cases of deviance, and with basic deviance it covers the most important and most troublesome kind of causal deviance. I take this to be sufficient to show that the problem of causal deviance can be resolved convincingly.
of nervousness that undermines the agent’s control occurs between the climber’s reason-states and that movement, which would be a basic action.
In the case of consequential deviance, the control-undermining states or events occur somewhere between a basic action and some action—or outcome—that the agent intended or wanted to perform—or bring about—by performing the basic act. Consider the following standard example. A sniper has the intention to kill an enemy by shooting him. He carries out the intention, but misses. By producing the noise of the shot, though, he stampedes a herd of wild pigs, which trample the poor enemy to death.45
In such cases the agent performs a basic action intentionally, for reasons, and the agent is in control as far as that basic action is concerned. The deviance occurs later and it affects an action or outcome that the agent wanted to bring about by
performing the basic action. These cases are thought to constitute counterexamples, because the agent’s end is not to perform the basic action. The sniper’s end is to kill the enemy; he intends to do so, and the intention causes the enemy’s death. According to a simple standard-causal theory, the sniper does not only perform the action of firing the shot intentionally and for reasons, but he also kills the enemy intentionally and for reasons. The former is true, but the latter, it seems, is clearly false. Hence, the standard-causal theory has got it wrong.
Cases of second-agent deviance are special cases of basic deviance, as control is undermined by interference between the agent’s reason-states and the movement that would be a basic act. Second-agent deviance occurs when control is undermined not by states or events, but by another agent. Harry Frankfurt has presented a much- discussed example, which fits exactly this description, in order to make a point about moral responsibility.46
The agent, Jones, is about to decide whether to vote for one of the two presidential candidates. Unbeknownst to Jones a gifted neuroscientist, called Black, has implanted a device in Jones’ brain by virtue of which he can detect what Jones is going to decide and influence his actions, if he wishes to do so. Jones, it seems, it not fully in control, no matter whether Black intervenes or not. And it seems that the causal pathway is deviant—at least in those cases in which Black intervenes.
45 Compare Bishop, 1989, p. 126, who has taken this example from Daniel Bennett. For another much
discussed example of that kind see Chisholm, 1966.
Let us consider first a standard way of dealing with consequential deviance. What goes wrong in cases of consequential deviance is that the intended action or outcome is not brought about according to plan. The sniper intends to kill the enemy in a certain way; namely, by shooting him, not by having him trampled to death. The standard-causal model can accommodate cases like that by requiring that actions are guided by the contents of the relevant reason-states. In cases of consequential deviance, the agent has a certain action-plan that is incorporated in the contents of the relevant reason-states. In the example, the action-plan consists in the intention to kill the enemy by shooting him. To say that the performance of the action must be guided by the intention is to say, simply, that the way in which the intended end is brought about must be in accord with the agent’s action-plan, which is incorporated in the content of the intention.47
The notion of guidance can be accounted for in causal terms, if it is granted that the intentional contents of reason-states and events can be causally relevant. Guidance by reason-states can then be construed as causation by reason-states in virtue of their contents. A bodily movement, for instance, is in that way guided by reason-states, because the reason-state’s content is causally relevant as to whether that particular movement occurs, rather than another one or no movement at all.48
Requiring guidance by reason-states, the standard-causal model can exclude cases of consequential deviance.49
47 For a summary account of the role of action-plans in intentional action see, for instance, Mele and
Moser, 1997.
48 Consider, for instance, guidance by an intention—by an agent’s having or acquiring an intention.
The simplest form of an action-plan, which is part of the content of an intention, consists of a representation of the intended action; a representation of which type of action is to be performed. In most cases, though, the action plan specifies by what means—by the performance of which type of act—the intended or desired end should be attained. An act-token is then guided by an intention if, firstly, it either instantiates the intended act-type or if it instantiates an act-type that is specified as a means to the intended end, and, secondly, if it is caused by the intention in virtue of its content—which incorporates the action-plan.
49 Many have argued that the causal connection between reason-states and the action must be sustained
or continuous—at least in some cases where the agent’s exercise of control is a process that takes as long as the performance of the action itself. Compare for instance Bishop, 1989; Mele, 2003; Thalberg, 1984; Searle 1983. The standard way of accommodating such cases is to say that the guidance function of intentions—or reason-states in general—can be sustained or continuous. That requires that the agent
is having the intention as long as the execution of the action takes—which seems unproblematic. The causal relation between such an intention and the action might be construed as sustained or continuous
causation. But it can, more plausibly, also be construed as a series of feedback loops between behaviour and intention. That means, very roughly, that the causal pathway would incorporate a mechanism that adjusts, at certain time-intervals, the execution of the action in accordance with the content of the intention. Compare for instance Bishop, 1989, pp. 170-171 and Mele 2003, pp. 56-58. Further, cases in which the agent must improvise cause complications—either because the plan is not
Basic Deviance and Reason-Responsiveness
To refer to the guiding role of the contents of reason-states helps only when the agent has an action plan that covers the event that would have been an action, had the pathway not been deviant. That is why it helps only in cases of consequential deviance. Consider again Davidson’s climber. He wants to rid himself of the weight and danger and believes that loosening his hold on the rope is a way of doing so. He does not have a further belief how to loosen his hold; he does not have a belief concerning what means are necessary, appropriate or best for loosening his hold, because loosening the hold is a basic act. We do not need to plan basic actions, because we do not have to know how or by what means to bring them about. We just do them; that is why they are basic. And that is why reference to guidance by reason- states does not help with basic deviance.50
It has been pointed out that an action that is done for a reason is a response to that reason; action that is motivated by reasons is responsive or sensitive to reasons.51 In the previous section I have outlined how the requirement of reason-responsiveness can help to handle examples of basic deviance. I will argue now, firstly, that the problem of basic deviance can in fact be solved by requiring reason-responsiveness, and, secondly, that the notion of reason-responsiveness is compatible with the standard-causal model.
The reason why the causal pathway in the climber example is deviant seems to be that the movement is merely caused by the reason-states—the movement is not a response to the reasons qua reasons. The relevant events in that example are following: the agent’s having the reason-state (say, the intention to rid himself of the weight and danger), the agent’s being nervous, and the agent’s movement of loosening the hold on the rope; call these events r, n and m. In the actual scenario r
causes n, and n causes m. What will be relevant to an explanation of why the causal pathway is deviant are the following three facts. Firstly, m is caused by an event, n, which undermines the agent’s control. Secondly, n does not rationalise m, but is caused by an event that rationalises m. And thirdly, the fact that m is caused and
rationalised by an appropriate reason-state, r, is a coincidence—because it is a
specific enough or because things are not going according to plan. Brand, 1984, and Bishop, 1989, show that the guidance condition can be refined as to accommodate such cases.
50 Compare Bishop, 1989, pp. 132-134.
coincidence that n causes the type of movement that is rationalised by m. Basic deviance is possible because coincidences of that kind are possible.
Proximate Causation
A first strategy to solve the problem of basic deviance is to require proximate causation. It says that intentions are the proximate or immediate causal antecedents of all actions, and it excludes control-undermining states by excluding all causal intermediaries between reason-states and actions.52
(Note that it would be problematic to exclude only control-undermining states and events, because we must explain agential control in terms of non-deviant causation.)
Bishop has dismissed that strategy for the following reason. Intentions, Bishop claims, ‘as realised in central neural states will never be causally proximate to the bodily movement that matches them’.53 There will be a causal chain of physiological events that results in the movement, and the most the proximity strategy can require is that the intention initiates that chain. But that, of course, does not solve the problem, because that causal chain may run through a state or event that undermines control.
That objection, however, is flawed as it confuses levels of explanation. The requirement that intentions must be proximate causes of action is a requirement at the level of psychological description and explanation. It may well be—and it is almost certainly the case—that a causal relation between two intentional events is realised at the neuro-physiological or physical level by a far more complex chain or pattern that involves a multitude of events. But that does not show that the intention cannot be the proximate mental antecedent of action.54
The proximity strategy, however, is unsatisfying in one important respect. I said that the problem in the climber case is that the bodily movement is merely caused by the reason-state, rather than being a response to it. The proximity solution ensures that bodily movements, for instance, are caused and rationalised in a way that constitutes agential control. But it does not show that the bodily movement is a response to the reason-state qua reason-state; it does not show that the reason-state causes the bodily movement because it is a reason-state. The reason-state causes and rationalises the action, but the reason-state’s rationalising the action seems to be irrelevant to its
52 Compare, for instance, Brand 1984, Mele and Moser, 1997. 53 Bishop, 1989, p. 139.
causing it. The proximity strategy establishes agential control by excluding all causal intermediaries—which excludes, trivially, all control-undermining intermediaries. But it does not establish control as reason-responsiveness.55
The Counterfactual Strategy
A second strategy for solving the problem of basic deviance attempts to capture the notion of reason-responsiveness in counterfactual terms. Let us consider again the climber example and assume that in the actual scenario the climber has the intention of ridding himself of the weight and danger by loosening his hold on the rope
now—at time t. And consider a possible scenario in which the climber intends to rid himself of the weight and danger by loosening his hold on the rope not now, but
shortly—at t’. Other things being equal, that intention would unnerve the climber just
as in the actual scenario, and we can assume that the climber’s nervousness would cause the loosening of the hold on the rope now—just as in the actual scenario. We can see, then, that the agent’s movement in the actual scenario is not reason- responsive, in the sense that the climber would not have performed a different action, had he had a slightly different reason-state—an intention that calls for a performance of the loosening of the rope at t’ rather than at t. Given that, it seems that an appropriate counterfactual condition can solve the problem of basic deviance. Consider as a first and rough approximation the following condition, taken from Berent Enç’s recent treatment of the problem of causal deviance.
(CC) If the content of the intention had been different, the action would have
been different correspondingly.56
The required correspondence between the intention’s content and the action consists in a match between the type of the basic act performed by the agent and the type of basic act specified in the content of the intention. So, had the agent intended to B, rather than A, then the intention would have caused a bodily movement that is, or constitutes, an action of type B (and had the agent intended to bring about the end E
by performing the basic act B, rather than by performing A, the agent’s intention
55 Another worry is that the proximity strategy is ad hoc, as it simply excludes what is responsible for
the problem. But there is independent plausibility to the claim that all actions are preceded and accompanied by the agent’s having or forming an intention. Searle, for instance, argues on