Friday, September 13, 2013

Hempel on the symmetry of explanation and prediction

To: Mark S.
From: Geoffrey Klempner
Subject: Hempel on the symmetry of explanation and prediction
Date: 10th November 2010 11:31

Dear Mark,

Thank you for your email of 30 October, with your essay for the University of London BA Methodology module, in response to the question, ''Whatever suffices for explanation suffices for prediction and vice versa.' Discuss.'

This is an excellent essay with which I can find very little to disagree. I think you've really hit the nail on the head with your explanation (!) of why counterexamples to Hempel's symmetry thesis arise in the first place; that although citing the cause of an event explains the occurrence of that event, there are cases where we are able to predict an event on the basis of an observation which does not reveal the cause of that event, e.g. because both observation and predicted event are results of a prior cause, and also cases where we successfully explain an event which we could not have reliably predicted. e.g. because our ability to identify the cause depends on the wisdom of hindsight.

Also, I don't find that I am in any disagreement with you regarding the irreducibility of our notion of a cause (given your previous essay and my comments, some of which you have incorporated into this essay). As a Humean (in the widest sense) about causation, I don't believe that I have any handle on the concept of causation other than the idea that things we call 'causes' and 'effects' are particular instances of the workings of laws of nature. This is consistent with holding that causation cannot, as Hempel's DN model claims, simply be reduced to deductions of instances from a covering law and initial conditions. In my previous email, I suggested that agency and our ability to intervene in the course of nature was a major component in the idea of causation. This is my explanation of the point you acknowledge in your essay, that the concept of a cause is embedded deeply in our conceptual scheme. As subjects of experience we are necessarily and not accidentally agents.

You talked very confidently about what is, or is not, an 'explanation'. I did wonder, however, whether at least part of the answer to this question involves something to do with the question why, as observers and also as agents, we are interested in explanations. If explanation is knowledge, what kind of knowledge is it? If it is a tool, what kind of tool? Are all explanations (as Lewises rather simple thesis seems to suggest) all basically of the same kind?

I'm thinking particularly of a thesis memorably stated by Hilary Putnam, originally in a series of lectures which I attended at London University, and subsequently published in his book, 'Meaning and the Moral Sciences', which is that explanation is 'interest relative'. Suppose you asked a criminal 'Why do you rob banks?' The criminal answers, 'because there's more money in banks than there is in gas stations.' Whereas, you were looking for an answer to the question why he is a criminal (e.g. 'Because I had a deprived childhood and fell in with a bad crowd').

This interest relativity becomes much more apparent in the 'moral sciences', in history, psychology, sociology etc. which surely are just as relevant areas of explanation and prediction as physics or chemistry. Arguably, even if we accept the interest-relativity of explanation, this leaves the link with causation intact. If your interest is reducing the number of persons who take up bank robbery as a profession, then the causes in question are those relating to social deprivation or the failure of prisons as a means of reform and rehabilitation, whereas if your interest is in reducing the number of bank robberies, then the causes in question relate to the lack of sufficient security which make banks attractive targets for criminals.

So far so good, but it seems to me that once we leave the paradigm of the core sciences and consider explanation in its wider context, it becomes progressively harder to maintain Lewises thesis. Why does Beethoven choose a cello to play a passage where you'd expect to hear a violin? Why does Pip in 'Great Expectations' risk his life to save Mrs Haversham? (Why, for that matter, am I having such difficulty thinking up good examples? Does 'I'm not feeling very creative today' involve some kind of implicit reference to causal history?)

Anyway, you get the general picture. 'Explain' is a concept which is put to a variety of uses, as indeed is 'cause'. But why should there be an interesting or simple account of how the two concepts are related?

What this shows is not that you are wrong in claiming an important link between causation and explanation, but rather that it is worth asking the question, Why does citing a cause or causes satisfy our desire for explanation in the way that it does? What more general phenomenon is this (identifying a causal pattern or relationship) a particular example of? An answer to that question would strengthen your case rather than weaken it.

All the best,

Geoffrey