Wednesday, December 18, 2013

Inference to the best explanation

To: Scott B.
From: Geoffrey Klempner
Subject: Inference to tbe best explanation
Date: 21st April 2011 18:28

Dear Scott,

Thank you for your email of 11 April, with your essay for the University of London BA Methodology module, in response to the question, 'Is inference to the best explanation just a special case of more general forms of confirmation?'

I enjoyed reading this. You have a knack of explaining philosophical ideas in a way which is gripping and intuitively appealing. Your general strategy in this essay seems to have been to cast about for as wide a variety of examples of 'inference to the best explanation' as possible, looking for ways in which these might support the idea that there is something different about the notion of inference to the best explanation, something that is not reducible to 'more general forms of confirmation', i.e. induction.

Your conclusion, however, is that the way we decide between competing explanations is, ultimately, based on inductive grounds even if this is not immediately apparent. Typically, we make assessments of the probability of competing explanations or hypotheses, and these probability judgements are based on more general knowledge which we have gained from experience and induction. You also talk about the 'regulative principles' which guide our selection of the best explanation, which themselves are ultimately justified inductively by the fact that they have worked in the past.

As you note, inference to the best explanation is a pervasive phenomenon. It certainly looks different from induction. One can understand the enthusiasm of philosophers of science such as Peter Lipton. But how important or novel is this concept, really? How does the observation that inference to the best explanation differs from the process of induction bear on the questions of knowledge and justification, which are the main focus of methodology?

The question you have been given is typical of a range of questions all based around the alleged 'difference' between inference to the best explanation and more general forms of confirmation. It is worth making the point that the *best* explanations are not always the most *obvious* ones. If you just reasoned by standard induction, looking for similar patterns, using concepts or forms of explanation which have worked in the past, you might never get there, never make the necessary breakthrough.

It's a story which has been repeated again and again in the history of science. But it is also the stuff of popular fiction -- the detective who refuses to go 'by the book' and instead chooses to think out of the box, his creative imagination supplying the link between scattered clues that did not make any sense to his more plodding colleagues.

But aren't we confusing two very different things here? Take the example of mathematics. Some of the more interesting theorems in mathematics required genius to discover them. But once discovered, the proof is a schoolboy exercise.

When we talk about inference to the *best* explanation there are two components: thinking up possible explanations, and then testing or judging their relative attractiveness or probability. The 'inference' part isn't in the creative generation of possible hypotheses or explanations. There's no procedure of inference which will give you an original explanation that no-one else has thought of before. The inference comes where we compare explanations, appealing to the kinds of considerations which you talk about in your essay, assumptions about probability, general inductive knowledge, regulative principles etc.

Once we have a range of explanations, there can be rational debate, where all the considerations put forward refer to or imply knowledge based on induction. This is the 'method' of science. But there is no methodology for generating explanations. Some people have a gift for this and others have to rely on familiar patterns of reasoning, analogies with explanations which have worked in the past and so on.

When Popper talked about the 'Baconian myth' he was thinking along the same lines: that you will never come up with original explanations merely by extrapolating familiar patterns. There has to be an element of creative imagination. There's no mystery about this. Dennett talks of the 'generate and test' model for creative thinking. Undoubtedly, traits of character play an important role -- such as intellectual courage, an iconoclastic tendency, refusal to follow the conventional route. All attributes that a good scientific education ought to foster. But these are not, strictly, part of methodology, testing, confirmation -- all the things that determine our degree of belief or confidence in a given hypothesis. Here, the only thing that we can fall back on ultimately are methods and procedures that are tried and tested, and through their success have given us the confidence to apply them to new cases.

All the best,

Geoffrey