To: Mark S.
From: Geoffrey Klempner
Subject: The justification of inductive reasoning
Date: 30th November 2010 13:24

Dear Mark,

Thank you for your email of 22 November, with your essay for the University of London BA Methodology module, in response to the question, 'What, if anything, can explain the rationality of reasoning according to inductive principles?'

In this very long (nearly 6700 words) essay, you have undertaken a wide-ranging evaluation of all or most of the significant responses to the problem of induction. Obviously, you could not do this in a one hour examination answer. However as in previous essays I accept that this has more than one purpose. The work you have done for this will provide adequate ammunition for a range of possible exam questions.

What would a good exam answer look like? When a question is as general as this, I think that you are entitled to state at the beginning that you will look at one or two arguments that you consider the strongest, after disposing of the others in as few words as possible.

As I read through the various responses: from Kant, Popper, Reichenbach and Salmon, Strawson and Edwards, Putnam, Van Cleeve and Papineau, my overriding impression was that each of these represented a 'solution' of a kind, not a dead end or an example of fallacious reasoning. Maybe that just shows something about my natural tendency towards eclecticism. However, there is something that you say right at the end of the essay which raises a warning flag for me: I suspect that you have been overly tough with these putative 'solutions' (to use a generic term which includes what you term vindication, validation, deflation, deflection and ad hominem retaliation!). You state: 'Frank Ramsey was probably correct when he wrote that asking for a justification of induction 'was to cry for the moon'.'

Why? Because 'we have no idea what the future holds'. It's that simple. The very next moment my G5 computer might turn into a fire-breathing dragon. I can't prove that it won't. But is that all Hume was saying?

As you state at the beginning of your essay, the point isn't just that inductive inference doesn't produce the same degree of certainty as deductive inference. Everyone accepts that. The point is that we believe, falsely, that performing inductive inferences is reasonable, and either that this can be shown to be the case, or that it is not necessary to show that this is the case.

So let's play a game of make-believe. One hypothesis would be that nature is not uniform. The little bit that we have observed appears to be uniform, but on a wider view (which we have not yet attained) it turns out that various degrees of chaos reign. The problem with this is that it is not the worst case scenario. As science continues on its way, the threat seems to recede. Indeed, there is always the possibility that we could discover that we inhabit an island of regularity in a chaotic universe (somewhat like Anaximander's cosmology of the Apeiron). But this is not the worst case scenario because things would be even worse for us if we were the playthings of a mischievous deity, who deliberately leads us down blind alleys.

I am going to assume that we do not and cannot prove that the worst case scenario doesn't hold, and see how the various solutions that you canvas fare:

1. Kant. Your example of the alien Connie who always reasons counter-inductively, and Randy who always spins a coin, miss their target here. Kant's argument in the Transcendental Analytic broadly seeks to establish that experience is only possible on the condition that induction is justified. If I can't validly perform induction, then there are no such things as spatio-temporal particulars. If there are no objects of my perception, then there is no 'I'. The mischievous deity cannot, logically, create a world where either Connie or Randy exist as self-conscious, experiencing subjects. To me, that's an impressive result. It doesn't go all the way (a point made by Strawson in 'The Bounds of Sense') but it constitutes at least a partial answer to Hume.

2. You say that Popper's 'conjecture and refutation' are insufficient as a basis for belief, and yet it is beliefs that are a necessary condition for action. Well, I'm going to define a mental state called schmeblief. In the absence of beliefs, we rely on schmebliefs. We boldly go where no-one has gone before. But do we? I guess you would say that if *any* theory is OK so long as it is falsifiable, not just OK to test and consider but OK to act upon, that would be very scary indeed. That's true, but testing is precisely what we do when we are unsure whether or not to act. The table looks a bit small for my G5, so I give it a bit of a wobble and a push and it feels solid enough. That's what I mean by schmeblief, and also what I think Popper is getting at with 'corroboration'.

3. Reichenbach's response seems rather similar in this respect. Life is a percentage-game, you can't always rely on playing a winner. 'Belief' is too clumsy a concept to capture this idea. My best bet is that the table will not collapse.

4. On Strawson's behalf, I would argue that it is part of what we mean by 'inductive grounds' that these grounds are defeasable. It is irrational to assume the existence of an mischievous deity, but of course we could be wrong. However, the rational thing to do is not assume the worst-case scenario, to do so would be paranoid.

5. I'm not sure whether it's OK to treat Putnam's argument for realism in science as an attempted justification of induction. It involves induction, to be sure, but it's aim is to combat scepticism about unobservables -- the view which motivates various anti-realist positions such as instrumentalism, fictionalism etc. Notwithstanding the observation that the history of science is a graveyard of discarded theories, surely it is still more rational to believe that our current theory is true than otherwise.

6. As a contextualist about knowledge, I'm not sure what to say about reliabilism, that hasn't been covered by the above. It has never occurred to me that a reliabilist would claim to have solved Hume's problem. We don't need internalist 'justification', the reliabilist will say. But it's still a valid question to ask what justification we *could* provide, if challenged.

All the best,

Geoffrey