Rationalist Interpretation of Causal Laws

There have been various attempts to rescue the intuition, discussed last time, that there is some sense in which causes actively cause their effects. If there is some such phenomenon, then there is a distinction between cause-effect relations and “accidental” associations. If we define cause and effect as nothing more than just constant association, there can be no such distinction.

To get a distinction between causation and accidental association, we don’t need the idea of an active process, the animistic idea, just the idea of a necessary association. If some constant conjunctions are necessary, we can take them to be causal, and reject other constant conjunctions, those that are not necessary as accidental.

In order to make this work, one needs an appropriate idea of what necessity. One can just say that it is causal necessity, but of course that is completely unhelpful, it just amounts to declaring that certain associations are causal and others not, with no additional explanation.

The rationalist interpretation of causal laws tries to supply a nontrivial answer to what sort of necessity is involved. There seems to be only one sort of necessity we know of that is clearly distinct from causal necessity, namely logical necessity. (If I’m wrong, and there are others, they can be used to generate theories analogous to the rationalist one, and the criticisms will be parallel.)

So, the idea is that it is logically necessary that a cause causes its effect.

How is that going to work? It must be that it is part of the definition of the kind of the cause, that it results in the effect. Example: it is part of what it is to be salt, that it is water soluble. Thus, whenever salt is put in water, it dissolves as a matter, not of experience but of logic---if it hadn’t dissolved, it couldn’t have been salt: we took the meaning of the kind “salt” to include solubility in water.

Question: You could get the properties of a kind wrong, couldn’t you?

It is a matter of experience how things behave, what is associated with which, and so forth, and so it, on this theory, is necessary to use that experience in determining what the appropriate kinds are.

To say this is to see how the theory is, while perfectly consistent, completely unhelpful. We just shift from skepticism about causes to skepticism about kinds: instead of doubting whether this salt will dissolve (salt does dissolve, there is, by definition, no doubt about that) we doubt whether this white stuff is salt. Same doubt, different language, no help.

As Ayer puts the point, we've simply moved the difficulty from determining whether c of kind C causes e to a new difficulty, that of determining whether c is of kind C.

How could anyone have thought such a thing might help? The answer is something like this: there is a problem about the relation between causes and effects, but it isn't clear what causes and effects are. If one just assumes that one knows, more or less, what causes and effects are, one gets to use the words unreflectively. The proposed solution is a solution to the original problem. If one just takes the words "cause" and "effect" at face value, one can propose the solution without being aware that the proposal involves shifting the meanings of the words. But it is only in the shift that the problem gets buried, and so, if you are paying attention only to the problem, not to the language in which you are describing it, you don't know where to look, and you miss that the problem is still there.

The reason I'm belaboring that point is that it is perhaps the chief insight of 20th Century analytic philosophy, one that is incorporated in essentially every bit of philosophy you will read or discuss in this department: We always, to at least some extent, engage in what Quine called "linguistic ascent": looking at the language which we are using in addition to using it to discuss whatever is being discussed.

There is no way to, in general, distinguish causation from constant association

Side remark: Outside of a few recondite stretches of philosophy, no one ever thought we were certain in that sense of causal laws. The fact that we don't have, for example, logical certainty about causal laws does not mean that we don't have good reason to believe that certain associations really are causal. The remark is not a skeptical one.

The first one is the doubt we had whether the arrow really was caused to fire by the tension. The second one is the doubt we had about whether the salt was really salt.

It seems to me that the argument is perfectly general, and that it shows that all of the currently accepted general characterizations of causation are empty, that is, no more informative than the brute, unexplained claim that some associations really are causal, while others are accidental.

Having said that, it is important for me to make clear that the argument only applies to the general case, not to specific theories. Just as Ayer denied that there are propositions and a general account of meaning while accepting that we can give specific accounts of meaning for specific "propositions," it seems to me that, in particular cases, we can and do have good reasons for making causal attributions.

In systems that we understand fairly well that are built to operate in accordance with a theory, we have good reason to take the laws of the theory to be causal laws. So, for example, part of the reason that turning the key starts the car is that electricity (a mark of causation no one has observed) flows through the starter motor.

We also have good reason to think that something was a cause of an effect if we act successfully to modify the effect by modifying the cause. Example: if someone has a fever and antibiotics cure it, that is good reason to think that the fever was caused by a bacterial illness.

We also have good reason to think that an association of one kind of event with another kind is a causal association if it figures in explanations of many, apparently unrelated phenomena. My favorite example is continental drift and turtle spawning.

The examples use rather different reasoning to conclude that there is a causal relation, and the marks of causation in the different cases are of quite different sorts. There are marks of causation, but there is no single kind of thing, a mark of causation, just as there are meanings of propositions, but there is no single kind of thing, a meaning.

Whenever you see a general, philosophical, wide-ranging account of causation, be suspicious: it isn't explaining anything. It is, at best, a literary variant on just saying, because its causal dammit. But that is not to cast doubt on whether we have good reason to think, in particular cases, that certain associations are not accidental.

My assessment of causal theories of knowledge

The argument for them is based on the idea that there is a usable general notion of causation. There isn't, and so the theories can't be any good.

Ayer doesn't reason that way. Instead, he concludes that there is a general notion of causation, namely, constant conjunction, and analyzes why constant conjunction isn't good enough for the purposes of the argument for a causal theory of knowledge. My terminology is different: I rejected causation, but I still owe you an explanation of why I don't think there is a good argument for a constant conjunction theory of knowledge. That bit is what I, in my way of putting things, take Ayer to finally show.

Note: for most purposes in this course, the official policy is to stick to Ayer's terminology when discussing Ayer in papers and on exams. The reason I'm departing for part of today, is that the kind of thing that I take to be an error that I've described is so persistent that I've come to believe, as a sociological matter, that the only way to extirpate it is to eliminate the word causation from philosophical discourse for the next few decades.

Ok. I’ve torpedoed, not only the idea of necessary connection as an account of causation, but also Ayer’s view that constant association is an adequate analysis. So, since Ayer is discussing a causal theory of perception, I have shown it wrong because incoherent. I should therefore supply a positive account of causation so we can see what Ayer and his opponents are talking about. Here it is: we take there to be certain empirical laws (of physics, psychology, whatever). They have two features: one, they are empirically good, and two, they are general. We can take a constant association to be causal just if it follows from our laws.

-- ShaughanLavine - 05 Mar 2003 - 01 Mar 2007