To: Mark s.
From: Geoffrey Klempner
Subject: Role of laws in scientific explanations
Date: 28th October 2010 13:01

Dear Mark,

Thank you for your email of 20 October, with your essay for the University of London BA Methodology module, in response to the question, 'What role, if any, do laws play in scientific explanations?'

This is a very thorough (and long) exploration of the question of the role that laws play in scientific explanations. But I am left feeling somewhat puzzled.

The source of my puzzlement, I would guess, is that like many philosophers (like Hempel for example, or like Russell) I do not really know what causation is. You acknowledge this as a challenge, in your section 2/1 'The problem of giving an adequate account of causation' but your conclusion merely seems to be to the effect, 'Why not take causation as a primitive notion seeing that any attempt to analyse causation runs into difficulties.' That may be so, but I don't know what I am being asked to take as primitive because, to repeat, I don't know what causation is.

You give five of the 'possible roles' that laws could play in explanations. Let me suggest a role which you don't consider: that explicit reference to laws is neither necessary or sufficient for scientific explanations, but all scientific explanations imply an implicit reference to laws. That is because scientific explanations of particular events are causal, and causation is, as Hume claimed, lawlike generalization. (It doesn't need stating that explanations of laws in terms of other laws involves an explicit reference to laws.)

Or perhaps, more accurately, the connections which we regard as 'causal' depend on a subset of lawlike generalizations, viz. those involving reference to the possibility (which may be realistic or imaginary) of human action and intervention in the course of nature. (Hence the ineliminable time element.) On this view, 'cause' is a concept which is of immense importance, it is central to the notion of human agency, to knowledge, to explanation. But it can be all these without being taken as primitive and unanalysable.

Hempel is wrong in thinking that all explanations can be reduced to the DN (or, DN or IS) model. The motivation for this analysis is clearly the desire to avoid having to talk of causation, as a consequence of his belief in the possibility of a Humean analysis. But it doesn't follow that if Hume is right about causation, then Hempel is right about explanation. On the contrary, the problems Anscombe raises about ceteris paribus conditions suggest that we can give a simple and complete explanation (e.g. of why the stone broke the window or why the ink stained the carpet) without being in a position to offer an adequate Hempelian covering law explanation. On the contrary, an explicit DN explanation will never be an adequate substitute for a simple statement that C caused E because of the ceteris paribus problem. Explanations are always general -- they explain classes of events -- but the things we want to explain, actual events, are particular.

Causation is directly observable. In many cases (as Anscombe states) we have much greater confidence that C caused E ('you gave me the measles') than in any explanation involving lawlike generalizations. That is just another reflection on the centrality of the concept of causation. The very notion of perception arguably involves (as Strawson has argued) a causal element, not in the discredited sense of Lockean representationalism but rather as constitutive of how we understand the notion of a perceptual point of view, and the necessary physical restrictions which that implies.

Counterfactuals can help to explicate our understanding of causation but (of course) they don't analyse it because the very notion of 'similarity' of worlds (as on Lewises theory) presupposes an understanding of causation. However, it does no harm to state that causation and counterfactuals are closely related; an understanding of one requires, and goes along with an understanding of the other.

I don't have much to comment on the arguments which you rehearse against the view that scientific laws are necessary or sufficient for explanation. I find the particular examples appealed to, like the flagpole or Henry the crow, unconvincing, because they are merely ad hominem -- directed against the orthodox DN model, when the real question we should be considering is the nature of causal relations as exemplifications of laws.

There is a broader question to consider: Why do philosophers ask the question, 'What is scientific explanation?' Practising scientists don't have any problem with identifying an explanation when they see one. Journal articles on physics or chemistry don't contain sections discussing whether a particular discovery, of a correlation between two phenomena, or a proposed theory to explain previously disparate observations, is an 'explanation' or not. The issue is whether the described experiment is reproducible, or whether the explanation offered is the only one available, or etc.

All the best,

Geoffrey