blog.kennypearce.net

Substances, Events, and Causes

Irreducible agent causation is quite a slippery notion. Many philosophers hold that it is not merely slippery, but unintelligible or incoherent. I take it that these philosophers have stated genuine problems which a proponent of irreducible agent causation needs to answer. However, in pressing objections to agent causation, philosophers sometimes make what seem to me to be pretty serious mistakes. First, sometimes they fail to include (explicitly) the qualifier 'irreducible.'* Second, they sometimes claim that the problem (or one of the problems) with agent causation is that it's a species of substance causation, and substance causation is unintelligible or bad in some other way. Third, they sometimes claim that agent causation (or substance causation in general) is radically contrary to commonsense and/or never thought of by the proverbial 'person in the street.' In this post, I want to argue that all three of these claims are false. Causation by substances (by which I here just mean things), including agents, is perfectly intelligible and is part of commonsense. Furthermore, the requirement of agent causation seems pretty clearly a requirement not just for free action, but for action at all. Nevertheless, there are serious puzzles about agent causation, and the puzzles get even trickier if one goes in for irreducible agent causation. In other words, what I'm claiming is that the real dispute should not be about whether there is such a thing as agent causation, but rather about what agent causation is and whether it is reducible (to event causation or something else).

To start with, take a case. Suppose I throw a rock and the rock collides with a window and the window breaks. Now, in plain language we don't go around throwing the word 'cause' in everywhere; more often, we use a word like 'make' or just an active verb. So suppose someone asks, 'what broke the window?' or 'what made the window break?' (Note that we usually talk about doing something to the window or making the window do something; we don't use event language explicitly even here.) This question could be answered by either 'a rock' or 'Kenny.' We'll say that a rock broke the window, or that I broke the window with a rock or something along these lines. In plain language, we are constantly citing things, including persons (agents) as causes. Now, we can also cite events as causes: you could say, 'it got hit by a rock' or 'Kenny threw a rock.' But typically, the substance-causal language sounds more natural.

Now, there's a perfectly ordinary, commonsense distinction, within this kind of substance-causal language, between agential and non-agential cases. (Although, of course, the word 'agential' is not part of plain language.) If one says 'Kenny broke the window' (or 'made the window break'), this imputes some kind of responsibility to me, it says that the breaking of the window was an act of mine, one for which I could (other things being equal) be held responsible. But no such imputation is made when one says 'a rock broke the window' (or 'made the window break'). There seems to be a particular causal role played by an agent which justifies this kind of imputation.

Now, of course, someone could, in principle, throw me at the window to break it. In that case, I would make the window break in just the same way a rock does, not in the way an agent does. So this isn't just a matter of different kinds of entities entering into the causal relation; these are different species of that relation.

Now all of this strikes me as perfectly straightforward and commonsensical. Furthermore, it seems that causing an event in the special agential way is a pretty good gloss (not a terribly informative analysis) on what's meant by acting. But we've got questions: what is causation? What is the relationship between substance-causation and event-causation? What is the special agential way of causing? These are all hard questions.

As philosophers, we tend to like parsimony, so we like to reduce things. If we could get down to one kind of causation, either substance- or event-, that would be nice. There is some reason to think that this can be done. Note that in 'normal' instances like the rock breaking the window, what happens can be described either in terms of one event (the collision between the rock and the window) causing another (the breaking of the window), or in terms of some substance intimately involved in the first event (the rock) causing the second event (the breaking). What this suggests is that substance-causal talk can ordinarily be translated into event-causal talk and vice versa. The question of which way the reduction should go is a tricky one, here as elsewhere, but there is at least one reason for favoring reduction of substance-causes to event-causes: it's much easier to give event-causal than substance-causal interpretations of the known physical laws. If the laws of physics are real important to one's metaphysical picture, then it makes sense to take an event-causal picture.

It should, however, be noted that this is a pretty weak reason. The known laws of physics mostly don't explicitly use the word 'cause' for the simple reason that they mostly don't use words; they are mathematical equations. To talk about the laws of physics as making causal claims (especially if these are understood in an anti-descriptivist way) is already to engage in metaphysical interpretation of physics. Furthermore, it is clearly not impossible to give an account of the laws of physics as substance-causal laws. This happens, for instance, if one holds (as Swinburne does) that the laws of physics are descriptive generalizations about the causal powers of objects.

So here's what we've got: both substance-causation and event-causation are part of commonsense. Substance-causation looks like it comes off better in plain language, but event-causation looks like it comes off better in physics (though perhaps not by much in either case). The most familiar and un-puzzling cases of causation can be described in either substance-causal or event-causal terms.

But here's the question: could there be an instance of substance-causation that didn't have an associated instance of event causation? Even if one reduces event-causation to substance-causation, rather than vice versa, there would still be something puzzling here. It's because of some fact about the rock (its position and momentum) that it caused the window to break. As soon as that fact about the rock came to be true, it did its causing, and that fact's coming to be true is an event. But the classical theory of agent-causation is supposed to be precisely such a case: it is not that the agent comes to be in some internal state and, as soon as she comes to be in that state, does her causing. It is not supposed to be in virtue of any state of the agent that she causes what she does; she just causes it. Agent-causation on the classical (e.g., Reidian) understanding is not supposed to be associated with an instance of event-causation at all. This is the really puzzling thing.

I think that everyone - even compatibilists, even Humeans - should accept that there is such a thing as substance-causation in general and agent-causation in particular. This is simply commonsense. The puzzles arise when we try to give a philosophical account of this and relate it to laws of nature, to event causation, etc., and to explain what distinguishes agent-causation from other species of substance-causation. On these issues, there are indeed some very pressing problems for the classical theory of agent-causation, and to these problems I will not now offer any solutions.

(In case you are wondering where this post came from, I just finished reading Randolph Clarke's Libertarian Accounts of Free Will. It should not, however, be supposed that all of my criticisms are directed at Clarke; he might well agree with some of them.)

*To be fair, I should note that many philosophers, implicitly or explicitly, use the terms 'agent causation' or 'substance causation' to mean 'irreducible agent/substance causation.' But part of what I'm trying to say here is that this is a mistake; we should first get clear about agent/substance causation, and then figure out whether it's reducible - after all, this is the procedure metaphysicians follow with most everything else.