1 CHAPTER 7 Agent Causation in a Neo-Aristotelian Metaphysics JONATHAN D. JACOBS & TIMOTHY O’CONNOR Freedom and moral responsibility have one foot in the practical realm of human affairs and the other in the esoteric realm of fundamental metaphysics—or so we believe. This has been denied, especially in the metaphysics-bashing era occupying the first two-thirds or so of the twentieth century, traces of which linger in the present day. But the reasons for this denial seem to us quite implausible. Certainly, the argument for the general bankruptcy of metaphysics has been soundly discredited. Arguments from Strawson and others that our moral practices are too deeply embedded in human life to rest on anything as tenuous as a metaphysical doctrine far from the thoughts of ordinary people would seem to prove too much: we can easily imagine fantastic scenarios far from the thoughts of ordinary people—involving, say, alien manipulation or massive deception—that, if true, would clearly undermine claims to freedom and responsibility. For still other philosophers, the separation of the moral life from (some) metaphysical issues is prescriptive, not descriptive: it is a recommendation that we revise ordinary moral thought by severing its allegedly problematic links to metaphysics. (Some philosophers appear to hover undecided between such a prescriptive project and a Strawsonian descriptive claim). We suspect that the prospects of retaining the binding force of ordinary moral thought, were such a reconceived moral practice widely embraced, are bleak. A transition to something closer to moral nihilism seems at least as likely. In any case, our interest here is in descriptive metaphysics, not revisionary. To say as we do that freedom and moral responsibility have a partly metaphysical character is not to suggest that they can be had only if some highly specific version of a particular metaphysical framework is correct. Instead, we suggest in what follows, it is a broadly neo-Humean metaphysics that is not hospitable to freedom (for reasons distinctive to the metaphysics), while a broadly neo-Aristotelian metaphysics is. But we 2 also think (and it is the main aim of our paper to show) that different versions of the neo- Aristotelian metaphysics lead to rather different metaphysical accounts of free and responsible action. Specifically, we will argue that (1) the most satisfactory account of human freedom within the broadly neo-Aristotelian metaphysics is agent-causal, but that (2) two different versions of the general metaphysics will lead to importance differences in the agent-causal account of freedom. Adjust the details of your general metaphysics, and the details of your account of freedom are transformed in significant ways. Action theory cannot properly be pursued in isolation from general metaphysics. 1. Freedom and neo-Humeanism David Lewis popularized a certain form of neo-Humean metaphysics, according to which causal facts and the laws of nature are reducible to facts concerning the global spatiotemporal arrangement of fundamental natural properties (which we allegedly may conceive in non-dispositional terms). Roughly, the laws are the best system of generalizations over such natural facts, where bestness is determined by the optimal balance of simplicity and ‘strength’ (explanatory power). Causation in turn consists in a restricted kind of counterfactual dependence of one event on another, where the counterfactuals are grounded in cross-world similarities.1 There are well-known problems with counterfactual accounts of causation, but we will not render any pessimistic verdict here.2 Furthermore, the problem that we do press against a neo-Humean account of free action is not dependent on a counterfactual theory of causation. It is a problem for any reductive account of causation, and we discuss Lewis’s picture simply for the sake of concreteness. 1 The locus classicus is Lewis’s article ‘Causation’, reprinted in his (1986). (We note that Lewis allows for temporally remote causation by defining causal chains in terms of stepwise counterfactual dependencies, but it is unnecessary to fuss about such details here). 2 For discussion, see the essays in Collins, Hall, and Paul (2004), which includes ‘Causation as Influence’, in which Lewis proposed a revision of his theory. Christopher Hitchcock (2001) and James Woodward (2003) have defended rather different counterfactual accounts that employ the structural equations framework that was given a major articulation and development by Judea Pearl (2000). For discussion of these developments, see Peter Menzies (2008). 3 Within the neo-Humean framework, intentional agency is naturally understood in terms of the counterfactual dependence of behavior or behavior-guiding intentions on appropriate beliefs, desires, or intentions the agent had immediately before and as the behavior occurs. That human beings act is (nearly!) uncontroversial. That we act freely can more plausibly be questioned. We assume here that both metaphysical freedom and moral responsibility are incompatible with causal determinism. Necessary conditions on free actions include plausible compatibilist constraints (e.g., the absence of strong internal or external compulsion) and that they are not determined to occur over some interval terminating in the initiation of the action. The inclusion of a non-negligible degree of indeterminism in one’s account of the proximate genesis of free actions is thought by many to give rise to problems of explanation and control. But questions of explanation and control are better posed within particular metaphysical frameworks. It seems to us that if the neo-Humean framework is accepted, indeterminism need not present a special problem of control. Causation is just counterfactual covariation of a certain kind, and the neo-Humean can readily describe a form of covariation of motivational factors and behavior that applies to the indeterministic case.3 (Indeed, this fact has been insufficiently recognized by compatibilists who have held that something approximating determinism is necessary for freedom). We should require only that the objective chance of the behavior’s occurring would have been much less in the absence of those factors. Furthermore, the counterfactual dependence of the chance of behavior on psychological facts with which the agent identifies is all that it could be for a person freely to form a choice. (Irreducible agent causation, for example, makes no sense in this metaphysics, so its omission can hardly be judged a deficiency.4) Hence, a suitably textured, causally indeterministic theory of free action gives everything that a neo-Humean could sensibly want for an 3 For an excellent discussion of this issue, see Randolph Clarke (1995). 4 Obviously, we are further assuming, though less contentiously, that whatever broad metaphysical account of contingent reality is correct for our world will hold for all worlds involving contingent concrete particulars. It is not the case that some worlds are neo-Aristotelian while others are neo-Humean. Without this assumption, the neo-Human account might well be deficient on grounds that there is a kind of direct control of action had by some possible agents though by no agents in a neo-Humean world. 4 account of metaphysical freedom (a fact that is insufficiently recognized by some agent causationists).5 We defer consideration of what is or is not explainable with respect to undetermined action, treating it in the context of our preferred libertarian account of freedom. It is true that, given a position held by some neo-Humeans (and others as well) that there are special explanatory limits in indeterministic worlds, there will indeed be a serious problem of explanation facing any indeterministic account of agency. But that position is not compulsory, and we will suggest below that it is implausible. In our view, the above neo-Humean account of free agency is founded on a deeply problematic general thesis of causal reductionism. By taking the fact of A’s being a cause of B to be a reducible, massively extrinsic relation—grounded in what occurs elsewhere and elsewhen—we empty the fundamental idea that causes ‘produce’ or ‘bring about’ their effects of any clear content.6 Since agency is a causal notion, this problematic consequence carries over: on a neo-Humean analysis, the sense in which my beliefs and desires here and now bring about my present action is at best very weak tea. A fortiori, extrinsic analyses, on which whether or not psychological factors are causes of behavior is metaphysically determined in large measure by what happens in the distant reaches of spacetime, provide a bizarre account of a free action’s being, as we commonly say, 5 Indeed, the extrinsic grounding of particular causal facts in the neo-Humean framework might lead one to doubt the necessity of indeterminism for freedom. See Beebee and Mele (2002). Unfortunately for the neo- Humean, this same extrinsicality renders it doubtful as an account of causation generally and of agency in particular, as we argue immediately below. 6 We should acknowledge that ‘causation’ in folk usage probably cannot be neatly lined up with a fundamental relation in the world, on any likely metaphysical account. The folk, for example, often speak of causation by absences, as when one says that Susan’s failing to water her neighbor’s plant caused it to die. On any plausible metaphysical account, there simply are no absences available to stand in a fundamental relation. In our view, it is most plausible to suppose that ordinary causal talk only roughly tracks an important fundamental relation in the world, which, to avoid contentious semantic disputes, we may call ‘M-causation’. The folk speak truly (often enough) even when speaking of causation by absences. But such truths are grounded in facts concerning ‘positive’ circumstances that stand in the M-causal relation (whose nature we sketch in the next section). Conversation with Gunnar Björnsson has helped clarify our own thinking here. There is also a nice discussion of this matter in Ted Sider’s Writing the Book of the World (2011), 15-16 and 75-76. 5 ‘directly controlled by’ the agent, such that it was ‘up to her’ what she would do in the particular circumstances.7 Our ordinary sense of control with respect to freedom of action manifestly points to something that supervenes on the local circumstances in which we act—or, at any rate, circumstances much more local than those thousands of years in the past or future. 2. A Neo-Aristotelian Metaphysics & Event-Causal Libertarianism There is more than one path away from the neo-Humean’s causal reductionism. Here we will consider only the path that we favor: a neo-Aristotelian metaphysics that assigns a central role to primitive causal powers. On this view, natural properties are, or of necessity confer, causal powers on their bearers.8 While the neo-Humean’s properties are intrinsically inert, the neo-Aristotelian’s are intrinsically powerful.9 If the neo-Humean world is ungoverned, since laws are merely descriptions of contingent regularities, the neo-Aristotelian world is self-governed, since laws are necessary descriptions of the powerful natures of properties.10 Neo-Humean causation is a sort of counterfactual covariance, but neo-Aristotelian causation is the exercise of an irreducible causal 7 See O’Connor (2009) for a development of this point. Gunnar Björnsson has pointed out in discussion that it is open to the neo-Humean to modify her account as follows: our concept of natural law require there to be some minimal score on the balance of simplicity and strength. In neo-Humean worlds where the patterns in one cosmic neighborhood or compact world-segment (such as the one we currently occupy) sharply differ from those in others, we should say that the laws themselves vary from one world segment to another. In this way, we needn’t say that what contingently occurs in very remote regions of spacetime are needed to fix what actions I bring about (or whether I ever so much as act at all). We grant that reducing the extent of extrinsicality serves to improve the view. But since it is the very nature of the view to give an extrinsic account of causation, this move cannot make the implausibility go away. There are ever so many neo-Humean worlds where memories and seeming historical traces are radically misleading beyond a short threshold into the past and where the patterns will abruptly change or simply cease in the very short future. Whether or not these things are in fact so just seems beside the point when we ask whether a present bodily motion is something that I freely bring about. (Note that the point concerns metaphysical determination, not epistemic justification). 8 See, e.g., Shoemaker (1980, 1998), Heil (2003), Mumford (2004), Bird (2007), and Martin (2008). 9 For full discussion, see Jacobs (2011). 10 See, e.g., Mumford (2004) and Bird (2007). 6 power.11 The details of the broadly neo-Aristotelian metaphysics need not concern us here. What are important are the general ideas, first, that properties are (or confer) primitive causal powers and, second, that causation is the exercise of such powers. In order to understand the nature of indeterministic causation within the neo- Aristotelian metaphysics, it is helpful to contrast it with another sort of picture that some contemporary philosophers endorse. On the latter, causal indeterminism is thought of as causation of probability. Indeterministic causes, no less than deterministic ones, are always and uniformly efficacious. They do not cause the undetermined outcome of an indeterministic process, but instead cause the outcome’s objective chance of occurring (generally by raising the probability, though in certain kinds of cases the probability may actually be lowered). Beyond helping to fix the prior chances of an event, there is nothing more that a cause does. Where the chance is 1, the cause suffices for the effect, and so we naturally, if misleadingly, say that it brings about the effect. But strictly speaking, nothing brings about the effect, whether the chance is 1 or less than 1. Only the prior chance is brought about. This causation of probability view is perhaps assumed (consciously or not) in objections to the intelligibility of indeterministic agency. If an outcome is not brought about by anything, it’s hard to see how it can be something that the agent controls and that we may fully explain in terms of her reasons for acting. But we should reject the causation of probability interpretation in favor of a probability of causation alternative precisely because the former makes the occurrences of events in indeterministic worlds utterly mysterious. There is no reason within a causal powers metaphysics to suppose that causes must always produce their characteristic effects, so that in indeterministic scenarios we have to resort to the fiction of regular causings of objective chances. We should suppose instead that indeterministic causes produce their effects though they need not have done so: they are propensities towards a plurality of possible effects. They are sufficient for each of them only in the sense that they are all that is needed, not in the sense that they are a causally sufficient condition.12 Every indeterministic event is produced, though none is necessitated. Causation, whether deterministic or 11 See, e.g., Martin (2008), Mumford (2009), and Bird (2010) for discussion. 12 Anscombe (1971) famously develops this point. 7 indeterministic, is a singular relation—the very same relation. The prior probability of one event’s causing another (with limit case of 1) is simply a measure of the strength of its (single-case) propensity to do so, which helps to fix applicable laws of nature. Let us apply this understanding to an indeterministic account of human free action that, like the neo-Humean account above, is rooted in a causal theory of action generally. According to it, when an agent freely acts, her web of motivational states is jointly disposed towards two or more choices, to varying degrees. Whichever choice is made, it will have been caused by some relevant motivation of the agent, a motivation with which she identifies.13 The exercise of agent-control consists in the causal efficacy of one’s motivations, and freedom further requires the openness of the future to (or consistency of the past and laws with) a plurality of specific outcomes. Such is a plain vanilla version of event-causal libertarianism.14 While agents, on this account, do not have any less control over what they do than agents in a corresponding deterministic scenario, they also do not have more. Indeterminism in the causal link between motivations and choice opens up a plurality of alternatives unavailable on determinism, but the agent does not seem to settle which of the options is taken in a sense robust enough for the agent to be morally responsible. Autonomous control seems to require more than compatibilist control plus plural alternatives. Consider two event causal libertarian universes, whose histories have been precisely the same until a time at which two intrinsically identical agents (including psychological propensities towards the same possible choices with the same degrees of strength) make diverging choices. It does not seem correct to say that it was up to the 13 The condition that the causing motivation be one with which the agent “identifies” is intended to handle possible cases where an agent might be subject to a powerful and perhaps momentary “alien” desire. We needn’t concern ourselves here with different accounts of this notion of “identification.” 14 The foremost recent defender of this theory, Robert Kane (1996), augments the account with further conditions on the process by which reasons result in choices. Laura Ekstrom (2000) locates the requisite indeterminism in a special subset of actions—those in which an agent critically evaluates her own conception of the good and comes thereby to have certain preferences that regulate ordinary actions. These proposals are interesting and it is worth considering the issues that they raise in their own rights. But they do not, in our judgment, suffice to answer the fundamental concern with event causal libertarianism that we raise immediately below. 8 respective agents, something that they were individually responsible for, that one chose the path of insult and the other that of gracious forbearance. It’s not that choices in these worlds would be ‘freakish’, the ‘result of pure chance’, and so not something that the agents in any sense did. It’s merely that the control that is exercised is of an insufficient variety to ground robust freedom and responsibility. But this objection presupposes the intelligibility of a stronger, more robust variety of control. Unlike on the neo-Humean metaphysics, there does seem to be space on the neo-Aristotelian account for such an alternative, as we will now show. 3. The Standard Agent-Causal Alternative Agent causalists maintain that freedom requires a distinct, enhanced kind of control from the causal efficacy of internal states with which one identifies. Responsibility-grounding control resides in an indeterministic, ontologically fundamental causation of a choice or action-guiding intention by the agent. Taking a feature as a metaphysical primitive is a reliable way to ensure that one's overall theory really does allow for the feature, instead of offering a pale substitute in the manner of various implausible reductionisms. And the longstanding difficulty of giving a plausible analysis of our pre-theoretical notion of autonomous control suggests that the gambit of primitive posit is not simply absurd. We should be careful to distinguish the agent causationist’s position from that of noncausalists (e.g., Carl Ginet 1990, Stewart Goetz 1988, Hugh McCann 1998, and Thomas Pink 2004). Both positions agree that autonomous control rests on a primitive capacity to form intentions (or volitions, according to the theorist’s preference). But the agent causationist insists that this capacity is—and can only be—causal in nature. The noncausalist, by contrast, ascribes ‘active power’ or ‘the power of choice’ to the agent while insisting that these terms are to be understood noncausally. However, it is unclear to us what this means. It seems to us that the term ‘power’ is being misappropriated for rhetorical purposes. Better that these theorists simply say that nothing causes free choices or volitions but that, notwithstanding, which choice is made is controlled by the agent, in 9 virtue of the fact that the choice is his.15 Such a statement is clearer—though clearly false, in our estimation. In assessing the agent causal account, we need to consider the role of the agent’s motivational states in the production of their undetermined choices. Randolph Clarke (2003) proposes an ‘integrationist’ account on which free actions are caused both by the agent (qua substance cause) and by certain of the agent’s motivational states (qua event cause). Clarke proposes that, in the presence of a ‘live’ agent causal capacity, it is a law of nature that: i) whatever action is performed will be caused by the agent, ii) a particular reason will cause an action only if the agent causes it, and iii) the agent will cause an action only if some corresponding reason also causes it. It seems to us that, absent further explanation, a lawful and symmetrical causal yoking of this sort is mysterious. Surely one or the other causal factor will be in the driver’s seat (given, as Clarke says, that they are not each partial causes). And we want there to be at least one sort of explanatory asymmetry: it is because the agent had those reasons that he (qua agent cause) caused the action that he did, not the other way around. Yet Clarke can’t say that the state of having those reasons indeterministically brought about the agent causal event on pain of making indeterministic causation by reasons more fundamental than (because prior to) agent causation in the production of an action. The resulting account would seem to offer no improvement over a simple event causal account that dispenses with primitive agent causation.16 O’Connor (2008) suggests a different account of the way that reasons influence agent causal actions. He suggests that while agent causal events are unproduced by other events, they are probabilistically structured by myriad factors, especially the agent’s own motivational states. As a result, agents have a continuously evolving, objective propensity to cause intentions to act in ways they take to be suited to their ends. More carefully, the 15 It is worth noting that defenders of libertarian accounts of freedom that are ostensibly event causal sometimes respond to the problem of control by emphasizing not that the choice is caused by the agent’s reasons but simply that it is ‘his’—it occurs within the agent. We think that the tendency of event causalists when pressed to shift between a causal and a noncausal, ‘ownership’ account of control is revealing. 16 For further discussion, see O’Connor and Churchill (2006). 10 idea is that motivational states act causally on the persisting capacity of an agent freely to form an intention to act, altering the objective strength of (or generating) the dispositions the agent has to form specific intentions within certain intervals. The influence of reasons so conceived is not unlike how things go according to the causation of probability interpretation of indeterministic causation. However, O’Connor’s account of the influence of reasons on agent-causal choices is not offered as an account of the nature of causation itself, and it does not have the absurd consequence that nothing brings about the specific outcome of an indeterministic process. This account has the advantage (over Clarke’s integrationist account) of offering a unified picture of the flow of causal influence, and it does so without sacrificing the core agent causal commitment to the exercise of a power that is not itself in turn produced by previous events. Note that on this neo-Aristotelian framework, having reasons (understood as motivational states) is having certain kinds of causal powers.17 So, as an agent first comes to have reasons for a course of action that was previously (subjectively) unmotivated, she comes to have new powers of choice and action. In a case where she comes to have additional reasons for an already motivated action type, the power so to act is not altered. However, insofar as the different reasons also motivate different actions of however fine grained a type, new powers are thereby acquired. And distinct reasons must have the potential to motivate somewhat different action types under at least some possible circumstances, for otherwise they could not be individuated within the causal powers framework. On O’Connor’s account of agent causal power, there is one persistent agent causal power, a power to form an intention to act. And an agent with that power can have differing specific propensities so to act, depending on what reasons the agent has. The reasons are, in part, powers to act on the persistent agent causal power, to alter its 17 ‘Reasons’ can refer to normative reasons, or the conditions (generally external to the agent’s psychological states) that rationally or morally justify a particular course of action for an agent in a given circumstance, whether or not the course of action is taken or the agent even acknowledges the existence of the reason. ‘Reasons’ can also refer to motivational reasons, the agent’s own reasons for doing what he does, wise or foolish as may be. In this latter sense, having a reason is a psychological state or set of states (such as beliefs, desires, and intentions) that motivates the agent towards and potentially explains certain courses of action. It is this latter, motivational sense of ‘reasons’ that is in view here.