Monday, March 12, 2007

Difficulties with Davidson's account of action (I): On explaining something in terms of the very causes by which it is defined

(by guest blogger Joshua Rust)

It is not uncommon to find philosophers defining otherwise problematic entities in terms of their causes. A plausible reading of Davidson, for example, sees him as defining action as a body movement which is caused by certain beliefs and desires. If the body movement is not caused by these attitudes, it does not count as an action. This is Davidson's answer to Wittgenstein's puzzle—when I raise my arm what is left over when I subtract the fact that my arm has gone up? (Philosophical Investigations, I-621) If the arm-raising is a genuine action, its causes are what remain following the subtraction.

Defining something in terms of its causes is not, of course, limited to philosophical discourse. Motion sickness is, perhaps, a certain queasy feeling that is caused in the right way. If one has the same feeling, but is on otherwise stable ground, that feeling does not count as motion sickness.

Here is my worry: if we define something in terms of its causes, it seems as though we are precluded from then explaining that thing by appeal to those same causes. This is manifestly clear in the case of motion sickness. If explanations are answers to why-questions, I may not cite motion in answer to the question, "Why am I experiencing motion sickness?" Why not? Because a condition of something's being an explanation is that it is genuinely informative. James Woodward, for example, says that a scientific explanation aims to "draw attention to further considerations the relevance of which is not apparent from [the why-question's] original characterisation of the explanandum under investigation" (Woodward 1979, 61; see also Mumford 1998, 139-41, and Braithwaite 1953, 320).

Davidson opens "Actions, Reasons, and Causes" with the following claim: "What is the relation between a reason and an action when the reason explains the action by giving the agent's reason for doing what he did? We may call such explanations rationalizations, and say that the reason rationalizes the action" (Davidson 1963, 685). But given Woodward's non-triviality requirement placed on a causal explanation, if actions are defined in terms of their causes (primary reasons = beliefs and desires), is Davidson then correct to say that a reason causally explains or rationalizes an action? That is, if (action = body movement caused by reason), we may not answer "Why (body movement caused by reason)?" by citing a reason. Reasons are not causal explanations of actions. This seems like a serious flaw with Davidson's account of action.

One possibility is that I've misconstrued Davidson's account of action; he does not define actions in terms of their causes. What is his solution to Wittgenstein's puzzle? Or what are the grounds by which he rejects this puzzle? Another possibility is that Davidson has something different in mind by explanation than does Woodward. But Davidson goes on to say "that rationalization is a species of ordinary causal explanation" (Davidson 1963, 685). As far as I can tell, actions aren't rationalized or causally explained by primary reasons—if anything, only a body movement is so explained. In the next post I will exploit this flaw to offer a non-standard reading of Davidson's account of action.

While this post is nominally about Davidson's account of action, I'm principally interested in leveraging the issue for metaphilosophical purposes. Any author which defines x in terms of its causes may not, if I am correct, then go on to claim that he or she has causally explained x by appeal to those same causes. Searle, for example, defines an institutional fact as a brute fact which has had some function intentionally imposed on it. He cannot, if I am right here, thereby go on to claim that institutional facts are causally explained by the said intentional imposition of function.


Anonymous said...

I'm not quite seeing the problem. Suppose that what makes behaviors actions, as opposed to something else, is that they are caused by belief-desire pairs. Now, take some particular action -- my raising of my arm. You're worried that we can't explain that action in terms of a BD-pair case, like Davidson wants to, because what it is to be an action is to be caused by a BD-pair.

But what exactly are we trying to explain? The behavior, which, it turns out, is an action. Why did he raise his arm? "Because of this BD pair." That is an informative explanation. Indeed, as Davidson says, to describe a behavior as an action is thereby to (partly) rationalize and explain it.

Maybe I'm missing something?

Joshua Rust said...

Jonathan--no doubt that we could give an informative explanation of a body movement (behavior) in terms of BD-pairs. As you said, in that case we will have learned something about the body movement--that it is an action. But we can't causally explain an action in terms of a BD-pair, because to be an action, by definition, on Davidson's view, is already to because caused by BD-pair.
If what Davidson is trying to explain is a body movement or behavior, his view is totally coherent. This is how you are reading him: "But what exactly are we trying to explain? The behavior..." Unfortunately, as the opening quote indicates, he thinks he is explaining actions and not body movements.
While he can explain behavior by appeal to primary reasons, he can't so explain an action.

I hope this is clear.

Anonymous said...

Some body movements are actions.

Take some particular body movement that is an action -- the raising of my arm. We want to explain it.

What's to stop us from explaining that movement by citing my BD pair? Sure, that BD pair is also what makes the movement an action. But what's wrong with that? It's perfectly natural to explain the movement of my arm by citing a BD pair.

If I've explained the movement, haven't I thereby explained the action? They are, after all, the same event.

Joshua Rust said...

Jonathan--thanks for sticking with me on this. Lets try to come to the same point from a slightly different angle.

I think we agree that the same event--say, my flipping the switch--can be brought under two descriptions:
1) it's simply a body movement.
2) it's an action. To be this, it must be caused by a bd-pair.
All actions are body movements, but not visa-versa (some body movements are reactions, spasms etc.).
Moreover, we agree that body movements may, in some cases, be causally explained by a BD-pair. What is in question is whether actions can also be so explained. Your question: "If I've explained the movement, haven't I thereby explained the action? They are, after all, the same event."

I don't think we can. If we know that the body movement is also an action, then we must, given the way we have defined "action," already know that the body movement is caused by a bd-pair (otherwise it is just a body-movement).
But if we already know that the body movement is caused by a bd-pair, then we can't go onto explain the action by appeal to the bd-pair. This is because, following Woodward, causal explanations must be informative--it must tell us something that goes beyond the scope of the explanandum.
We could causally explain the action, but for that we would have cite the causes of the bd-pair which is already known to give rise to the body movement (childhood events, etc.)
The point trades on non-controversial pragmatic claim about the nature of our explanation practices--namely that they must be informative. The implication is there are principled reasons to think that not all causes of an event can further function as a causal explanation--namely, when the event about which we are asking "Why?" is already known to be caused in a certain way. Actions, by definition, are already known to be caused by a bd-pair so that bd-pair can't further be a causal explanation (it violates the informativeness requirement).
This was the point I tried to make about motion sickness. Motion sickness is a feeling already known to be caused by motion. We can ask for a causal explanation of motion sickness. But, because of the pragmatic constraint, we can't causally explain motion sickness by appeal to the motion itself. We would rather have to cite the causes of that motion (like, air turbulence or being in an airplane).

Brad C said...

Very interesting post.

Devil's advocate:

When we claim a bodily movement, e.g. my fingers hitting the keys on the keyboard, is an action, we are committed to the claim that there is some BD pair that caused the action (as Davidson points out "in the right way" - no wayward chains). But we can do so without claiming any specific pair did the causing. We are just committed to the existential generalization being true, not that some specific pair makes it true.

When we explain the action by citing the specific pair we consequently do give further information.

ROUGH Example: Jane asked Bob a hard question. In saying she performed that action, I am claiming that there is some BD pair that caused her behavior (her uttering certain sounds).

But I might not be aware of the content of those mental states. I believe she had a pro-attitude towards acts of a certain type and believed that her asking Bob a hard question is an instance of that type, but that is all I believe.

If you tell me she asked the hard question because she wanted to embarrass him I now know the content of the desire and belief in question and have further information. I did not need to believe that info in order to think and claim her action was intentional, or even to do so responsibly.

Brad C said...

And I like the metaphilosophical point and am interested to see you tackle the part whole cases Lewis discusses (the murder of archduke Ferd. explains WWI even though it is part of WWI).

Joshua Rust said...

Brad--Excellent! You've very much anticipated my discussion in Part II! I think you are exactly right--citing a specific db-pair is informative. You might even say that it is explanatory. But don't think that it is casually explanatory for reasons that may already be obvious. So what is the force of the "because" that links a reason and action? Stay tuned!
Also thanks for connecting the discussion to Lewis' point.

Anonymous said...

I'm really not seeing this, Joshua. You say:

But if we already know that the body movement is caused by a bd-pair, then we can't go onto explain the action by appeal to the bd-pair.

This seems doubly wrong. First, it seems to assume that we can't explain things by citing things that we know. But of course we can. What else would we use to explain them? Second, it doesn't follow from our knowing that something was caused by a BD pair that we know that it was caused by the BD pair we cite in our explanations.

Clark Goble said...

Sorry - coming to the discussion late.

But does anyone really think that reasons as causes can be dealt with independent of Davidson's anomalous monism? After all it seems to be what grounds the issue. (I'd note that I see a very similar approach in Peirce)

Joshua Rust said...

Jonathan--your second point can be understood in two ways.

First reading: flipping the switch is an action (as opposed to a muscle spasm) because it is caused by some (any) bd-pair (we need not know which pd-pair in order to know that it was an action); but then a bd-pair is cited that specifies the content of bd-pair giving rise to the action. Wouldn't the latter be an explanation (after all, it's informative)? This is similar to Brad's point and will be addressed in part II.

Second reading: we know that flipping the switch is an action because it is caused by my wanting to turn of the light (a specific bd-pair). We can then ask, further, "why did he want to turn on the light (which caused the flipping of the switch behavior)?", and answer that question with a different bd-pair: because he wanted to read. This is totally fine: desire A can cause desire B which causes an behavior. Searle makes something of the same point when he distinguishes intentions in action from prior intentions. But I don't think this the kind of case Davidson has in mind when he asserts that reasons causally explain or rationalize actions. He is presumably making a point about the relation between behavior (what he calls "actions") and reasons. He is not, I think, making a point about the relation between prior reasons and the effects of less prior reasons.

Regarding the first point, the claim that explanations must be informative can't be argued for. If "because of the motion" seems like a perfectly good way to answer the question, "why does he have motion sickness?" then we have found a juncture at which our intuitions have parted. In my view the minimal informative-requirement makes it the case that not all causes are candidates for causal explanations; but you seem to want to read explanation and cause as roughly equivalent. This seems like a linguistic dispute (chose whatever word you are comfortable with to get at what I am here calling "explanation")--it's helpful that Woodward, van Fraassen, etc. share my intuitions, but there may be room for disagreement. On my view explanations always aim to fill a gap in our understanding; not all causes do this (only as a kind of joke is the response "I was born" considered an answer to the question, "why did you write this blog entry?"). I've tried to come up with a case where, on this definition of explanation, there a principled reason for thinking that certain causes may not be causal explanations--namely when the cause is embedded into the very idea of the thing explained.

Clark--let me think about your suggestion and get back to you.

Anonymous said...


Do you agree with me that if a BD-pair causes some behavior, it may explain that behavior?

If you agree with that, then you're in an awkward position if you're to deny that the BD-pair cannot also explain the action. Because the action is identical to the behavior.

I just don't see anything wrong with a bit of behavior being explained by the same thing that makes it an action.