O’Connor’s argument for indeterminism

Timothy O’Connor has recently defended a version of libertarianism that has significant advantages over similar accounts. One of these is an argument that secures indeterminism on the basis of an argument that shows how causal determinism threatens agency in virtue of the nature of the causal relation involved in free acts. In this paper, I argue that while it does turn out that free acts are not causally determined on O’Connor’s view, this fact is merely stipulative and the argument that he presents for this conclusion begs the question.


Introduction
In recent years there has been an attempt on the part of some metaphysicians of freedom to defend agent-causal theories of libertarianism. 1 O'Connor (1995O'Connor ( , 1996O'Connor ( , 2000O'Connor ( , 2002 has defended one such version of libertarianism, a version which he claims successfully bypasses the main criticisms which have been raised against such views. In addition, he claims that his view has the further advantages of: (1) Securing indeterminism on the basis of an argument that shows that an agent cannot be causally determined to act given the nature of agency and causation and, (2) Locating the indeterminism of a free action in a place that avoids the criticisms of alternative theories of libertarianism.
In this paper, I will focus on the first of these two advantages claimed for the view. In particular, I will argue that while it does turn out that free acts are not causally determined on O'Connor's view, this fact is merely stipulative and the argument that he presents for this conclusion begs the question.

O'Connor's metaphysics of properties and agency
Let me begin by setting out a few preliminaries in O'Connor's account, preliminaries that will play a role in my critique below. According to O'Connor, the agency theorist must begin by distinguishing between two sorts of properties that underpin the causal powers of substances. The first are "event-cause-enabling" (ECE) properties. ECE properties are such that when an object possesses ECE property P in circumstances C, having P in C necessitates a certain effect, E. So, possessing the property of "having the mass of 100 kg" gives a stone the power to crush an egg on which it is dropped. This property enables the rock to be the cause of the egg's being crushed. And, having that property necessitates the egg's being crushed when the egg and the rock bearing the property are properly situated (i.e. the rock with the property "being dropped onto the egg"). The second sort of property is "agent-cause-enabling" (ACE) properties. ACE properties do not necessitate a certain effect when placed in the appropriate circumstances, C, but instead such properties "make possible the direct, purposive bringing about of an effect by the agent who bears it," when the agent is in C (O'Connor 1996, 145). As a result, ACE properties simply enable an agent to perform a certain free act, while not necessitating that the agent possessing the property bring about that act. It is, then, possession of ACE properties that provide agents with the causal powers requisite for acting freely.
O'Connor then turns from a discussion of the nature of the properties which underpin the causal powers of the free agent to the nature of the free act itself. Earlier versions of the agency theory, such as those proposed and defended by Chisholm (1966, 12) and Taylor (1966, 109, 111 -115), characterize the act of the agent as the first event resulting from the agent's undertaking to act. Thus, the agent's act is to be identified with the first event in the chain of events leading up to my, say, raising my hand. We might call this first event the primitive or basic action to distinguish it from the whole causal chain of events involved in my hand being raised. O'Connor argues that agency theories of this sort misidentify the agent's primitive act. Instead, he claims, there are significant advantages to be had in relocating the primitive action from the first simple event, e, to the causally complex event "S's-causing-of-e" (where S is the libertarian free agent) (O'Connor 2009, 195;cf. O'Connor 2000, 72n11). On this view, "rather than there being a causal relation between agent and action, the relational complex [S's-causing-of-e] constitutes the action" (O'Connor 1995, 182).
Finally, O'Connor utilizes Dretske's distinction between structuring and triggering causes. According to this distinction, a structuring cause is one that simply arranges the causally relevant furniture in order to allow the causal powers of a substance to be exercised. By pushing the doorbell I am structuring the wiring of my house (by closing the circuit) in such a way as to bring about the event-causal complex of "the electric current's charge causing the doorbell to ring" (an event-causal complex is a causally complex event where both of the relata are events). I am not doing anything to bring it about that the power was generated, nor that it is being delivered to the house. I am just rearranging items to allow the charge to do what it does when it reaches properly functioning doorbells. 2 Triggering causes on the other hand, actually bring about the existence of whatever it is that is the front-end relatum of a causally complex event. There is, here, an ambiguity in the sense of the phrase "bring about the existence of" that is important for some points I raise below. Before we get to that, however, we should first outline O'Connor's argument.

O'Connor's argument for indeterminism
One of the most interesting advantages O'Connor claims for this relocation of the agent's action is that it provides him with an argument for the claim that prior events cannot deterministically cause agents to act. Given that O'Connor holds that all free acts are necessarily agent-caused and that human agents perform some free acts, the incompatibility of agentcausation and determinism entails indeterminism. 3 The argument hinges on the fact, he claims, that one simply cannot make sense of causally complex events as a whole, such as "S's-causing-of-e," having sufficient triggering causal conditions (STCC). O'Connor (1996, 147) puts the argument as follows: [It is clear] by the following simple argument that . . . an agent-causal event couldn't be causally determined. For instantiations of causal relations (causally complex events) are not themselves directly on the receiving end of other causal relations -instead, instantiations of intrinsic properties (causally simple states or events) are. Causal relations are the producings of events, rather than what are produced (in the first instance). Compare an ordinary case of an eventcausal process's (consisting of event F's causing event G) being caused by some further event, E. Surely this can consist only in E's causing F, the front-end relatum of the complex event. When I reflect on the matter, I cannot but regard this as reflecting an evident, general truth about causation. If this is right, then an agent-causal event could not be caused for the simple reason that the cause in this case is not itself an event.
The main point of this argument can be illustrated with an example. Consider a causally complex event in which the relata themselves are both events (ECC's). We can conceive of STCC for ECC's since to bring about the ECC, one need only bring about the "frontend" relatum (there is some ambiguity in using the phrase "bring about" here that I discuss below). So, for the ECC "the collision of the cue ball and eight ball causing the eight ball to move" we have the collision of the balls and the motion of the eight ball as our two relata, and we have the ECC, the latter of which can be causally necessitated by simply bringing about the collision of the balls. Here we can imagine STCC for the "front-end" relatum that then causally necessitates the "distal-end" relatum, thereby allowing the front end to provide STCC for the ECC in question. But there is nothing besides STCC for the front-end relatum that can act as STCC for the whole ECC. O'Connor then goes on to claim that the only way in which we can conceive of STCC for a causally complex event is in terms of STCC for the front-end relatum. O'Connor then concludes the argument with the line: "If this is right, then an agentcausal event could not be caused for the simple reason that the cause in this case is not itself an event." But if that is correct, says O'Connor, it is clear that agent causation precludes causal determinism. The reason is this. We should not regard the acts of agents as simple events but as causally complex events (call complex events of this sort "agent-causal complexes" or primitive actions). Unlike ECC's, primitive actions have as their relata not events, but an agent and an event, that is, S's-causing-of-e. But since the only sort of STCC we can imagine for any complex event are those which are STCC for the initial component of the causal complex, it is clear that we cannot have STCC for primitive actions, since it is hard to imagine what it would mean to bring about the agent that serves as the front-end relatum of the primitive action "S's-causing-of-e". Thus, on this view, says O'Connor, the very nature of the agent's free act precludes causal determination.

Unpacking the argument
It is unclear exactly how the argument is supposed to be understood. O'Connor wants to show that there is something about the nature of the primitive action, as he characterizes it, that precludes it from being caused (and thus causally necessitated). That is, there is something about the structure of a primitive action that precludes the possibility of there being sufficient triggering causal conditions of the action. But what exactly is it that precludes this? On the one hand he says that it is the fact that the primitive action is an instantiation of a causal relation, and "instantiations of causal relations (causally complex events) are not themselves directly on the receiving end of other causal relations." On the other hand he indicates that causal relations can be caused, but only in that sense that one brings about the front-end relatum of the causal relation which is in turn sufficient for the obtaining of the "distal-end" relatum (and so sufficient for the obtaining of the causal relation as a whole). But since, in the case of a primitive action, the front-end relatum is an agent, the relation cannot be caused in this way, since only events can be caused.

Agents, events, and causes
Above, triggering causes were characterized as causes that bring about the existence of the front-end relatum of the causally complex event. Of course, when we speak of causing the existence of the front-end relatum in cases of triggering let's notice that there is a measure of ambiguity in the notion of triggering cause. The ambiguity arises from the fact that one way of thinking about triggering causes makes them appear to be structuring causes in another guise. To see why this is so, let me note two ways in which one might think about triggering causes (in what follows I will speak of STCCs rather than just triggering cuases, since it is such causally sufficient conditions that O'Connor thinks is precluded on his account).
First, one might think of STCCs in terms of a set of causal conditions that jointly suffice for the actualization of a certain causal power of a substance. Notice this is not a case of activating a causal power of a substance but of actualizing it, that is, giving the substance a causal power it did not formerly have. But surely it can't be this that O'Connor has in mind here. In fact, it is hard to know what a triggering cause would be on this reading. Consider a case where I walk into my kitchen and notice that my partner has lit the stove but simply forgot to put the pot on the burner. I grab the pot, put it on the burner, and in a few short minutes the water is boiling along. It is clear, in this case, what the relevant structuring cause is that brings about the ECC of the fire's being hot causing the water's boiling, namely my putting the pot on the stove. But what would the triggering cause for the water's boiling be on this understanding of triggering cause? It could only be whatever it is that brings it about that the fire has the power of heating rather than not. But if that is what we have in mind, then it looks as if there either cannot be triggering causes ever or that the list of triggering causes is limited to some divine being or a very complex set of naturalistic conditions.
The other understanding of triggering one might have (and no doubt the one that O'Connor has) is that triggering simply amounts to bringing about the existence of the front-end relatum. On this view, we can imagine STCC's in the case of ECC's. Consider a case in which the pot of water is sitting on the unlit stove burner. I come along and light the burner, causing the water to boil. In this case I act as the triggering cause of the ECC by causing the existence of the flame.
But on this reading, why are STCC's precluded for primitive actions? There are, I think, two routes that O'Connor might take in holding that STCC's are so precluded for primitive actions. First, he might respond by pursuing the second reading of the original argument above. That is, he might say that the reason STCC's are precluded in the case of primitive actions is simply that the front-end relatum is an agent, and it is only events that can be caused. Second, he might respond that it is absurd to think that I can have it my power to bring about a primitive action by acting as the triggering cause in this sense. The trouble is that one cannot bring about the front end of a causal relation for which one is the front-end relatum.
What about the first response? Can S be on the receiving end of a causal relation? That depends on how we are to understand this front-end relatum. In the causal complex S'scausing-of-e, does S stand in for "S exists" (an event)? Surely we can think of triggering causes for that. Thus, O'Connor cannot take this to be the correct understanding of the front-end relatum.
But if that is not what is at work here, what could be? O'Connor could assert that S just means S (a substance or continuant). It refers to the agent and not some event or state of affairs that include the agent. But if this is the correct reading it is hard to see why it is any easier to conceptualize non-events as causes than as the results of the activity of causes. In other words, why is it an obvious truth to O'Connor that only events can be caused, but not that only events can be causes?
In any case, this second understanding of the front-end relatum will clearly not do. It is too strong to say on these grounds that no primitive action can have STCC's. The reason is that even on this understanding we can conceive of some triggering causes (that is, causes of the front-end relatum) for primitive actions, it is just that those STCC's cannot involve S (the front-end relatum). They might well involve, however, the parents of S, or God, or some complex set of naturalistic conditions, etc. Thus, we can conceive of triggering causes for primitive actions even if we take the front-end relatum of the causal complex to be a substance.
But what O'Connor should say in response to this claim is clear. To this criticism he should respond as follows: Fair enough -but even if this is right it is still true that causal determinism fails in the case of primitive action's since bringing about the front-end relatum here, that is bringing about the agent, simply isn't sufficient for the occurrence of the "distal-end" relatum, and so not sufficient for bringing about the causally complex event. 4 This seems right of course. But the same could be said for ECC's. Lighting the stove alone is not sufficient for bringing about the heating of the water (at least until some structuring goes on). And so, whatever role triggering has in one case (for ECC's) it seems to have in the other (primitive actions). But, one might think that all of this only serves to make the real difference between ECC's and primitive actions clear. In event causation, triggering the front end and structuring the world in a certain way is sufficient for the occurrence of the entire causal relation (cf. O'Connor 2000, 71). But this is not sufficient in the agent-causal case. And that is the real difference.
But now one is led to wonder why this difference obtains between the event and agentcausal cases. The only answer that O'Connor seems to have available is that the nature of the properties that allow for agent causation simply do not permit sufficient causal conditions at all. That is, the nature of ACE properties is such that their bearer is merely "enabled to purposively bring about some event e." And as noted at the beginning, this is to be distinguished from ECE properties just in virtue of the fact that in the latter case, having the property necessitates the effect in the relevant circumstances.
If that is correct, however, the real reason that primitive actions cannot have STCC's is that ACE properties have simply been defined in such a way that they preclude STCC's. In other words, rather than their being an argument for indeterminism here, it turns out that indeterminism is merely stipulative. 5

Conclusion
O'Connor wants to argue that if we conceive of the properties and powers that underlie exercises of free agency correctly, we get indeterminism (and thus libertarianism) for free. His argument for this relies on the claim that there cannot be sufficient triggering causal conditions for primitve actions because agents (the initial component of a primitive action) cannot be caused (i.e. there are no sufficient triggering causal conditions for agents). As I argue, it is not entirely clear that we can explain triggering causes in a way that applies to events that does not also apply to agents. Thus, the only way to account for what precludes determination in the case of agent-causal complexes is that we have defined agent-causal complexes in such a way that precludes such determination. The upshot of this is that there seems to be no credible move from O'Connor's theory of causation to agent-causal exercises being indeterministic in a way that is relevant for freedom-grounding control (cf. Markosian 1999Markosian , 2012.
Without a doubt, none of this impugns O'Connor's claim that on his version of the agency theory free acts will turn out to be incompatible with causal determinism. But it does undermine his claim that relocating the primitive action of the agent provides one with a conceptual argument for indeterminism. 6

Disclosure statement
No potential conflict of interest was reported by the author.
Notes 1. Some sample of the recent literature includes: Clarke (1996), Griffith (2007), Mawson (2011), Rowe (2006), Steward (2012), Lowe (2005), O'Connor (2011), and Jacobs and O'Connor (2013). 2. This is actually a slightly different rendering of the distinction than Dretske himself gives. O'Connor (1996, 157n13) introduces Dretske's distinction as follows: Fred Dretske (Explaining Behavior, Cambridge, MA: MIT Press, 1988, Ch.2) notes that there are two senses to be given to the idea of something's causing a complex process of the form, F's-causing-G. Consider the example of my pressing your doorbell (F) resulting in the ringing of the bell (G). Had the button not been wired to the bell in the appropriate way, (G) would not have obtained. The electrician who installed the device, it seems, was in some sense causally responsible for F's causing G, even though he did not in any way initiate this event. In Dretske's terminology, the electrician's activity was the structuring cause of F's-causing-G (by setting up the circumstances in such a way that F, should it occur, would result in G), though not its triggering cause (the cause of F, my pushing the button). But this distinction is not relevant to Ginet's claim, which clearly concerns the idea of an event's being the triggering cause of an agent-causal event.
One might think that this is the sort of thing that medieval concurrentists had in mind when describing God's relationship to creaturely causation in premotion. On this view, God acts as the triggering cause by virtue of the fact that he is responsible (i) for the fact that substances have the causal powers that they do (by way of his creating and conserving them) and (ii) for the fact that on a certain occasion the causal power of a substance is exercised only in conjunction with divine premotion. Aquinas, for example, provides as an analogy for (ii) the activity of the woodsman who occasions the exercise of the cutting power of the ax when he uses it to chop wood (cf. De potentia Dei Q.3, a.7).
Let us simply define a minimal notion of ability, such that one is able to make it the case that either p or not-p in this sense just in case it is open to one so to act (reliably or not) that it might be the case that p, and open to one so to act that it might be the case that not-p.
Here, the nature of agency (or, at the very least, the nature of agency enjoyed by rational creatures) precludes determination of any kind. For O'Connor, the powers of rational agents are such that they are necessarily undetermined (hence, his notion of "ability" is defined indeterministically).