To: Julian P.
From: Geoffrey Klempner
Subject: Possible worlds and counterfactuals
Date: 19 April 2004 15:45

Dear Julian,

Thank you for your e-mail of 7 April, with your University of London essay, 'Do Possible Worlds Provide a Satisfying Account of Modality?'

I find myself in the position of agreeing with you that we evaluate counterfactuals by constructing models, but also wanting to save possible worlds.

Let's start with the second point first.

I agree with you that similarity of possible worlds is useless as a concept for evaluating counterfactual statements. There are too many cases where it gives the wrong answer. It is too context dependent.

When you look at how we actually go about assessing the truth of counterfactuals, the crucial elements are time and causality. We look at some temporally antecedent condition which might have been changed, then use our knowledge of the way the world works to decide what would have been the result. In the counterexample 'If Oswald had not killed Kennedy someone else would have,' the Warren Commission supporter considers how the world would have turned out if, say, Oswald had been prevented at the last minute from firing his rifle, and not how things are in the most similar world where Oswald does not fire his rifle.

The reason we do this is connected with our interest in making counterfactual statements, the point of having such a locution in our language.

Having said that, it must be acknowledged that counterfactual statements can be vague, pointless, impossible to assess and people generally accept this. (Hence the saying, 'if all the world was apple pie, and all the sea was ink...'.) So is there any real value in insisting that counterfactuals must somehow be shown to have truth conditions? An alternative, anti-realist view is that a counterfactual statement is not a 'statement', or the assertion of a 'proposition' but a condensed argument. The clear cases we can call 'true' or 'false', but all we mean by that is that we are persuaded or unpersuaded by the hypothetical situation put forward for appraisal.

Now, Possible worlds are put forward as a way to justify the attribution of truth conditions to counterfactuals. At least, that is how David Lewis approaches the question. The reasoning here is that if we are prepared to attribute truth conditions to a locution, then we are committed to whatever ontology is required to account for those truth conditions.

In the case of possible worlds, the idea is that even though they might seem rather dubious entities, their existence is logically entailed by something else we believe, and are committed to.

I don't buy this because I think there are other, stronger reasons for believing in possible worlds.

My strategy would be to try to persuade someone who claims not to believe that there are 'ways things might have been' (as you declare on page 4) that it is in fact impossible to consistently hold this view. We construct models in order to predict the future, e.g. the weather. But we also use models in making sense of the world as a world of opportunities and threats, a world where people do things which deserve praise or blame. Thinking about these notions essentially involves thinking about what *might have been*. (Perhaps you do agree with this, because at the end of your essay you say that there is 'no analysis of modality'.)

I agree, Lewis can equally be criticized for effectively doing away with modality. When I kick myself for carelessly losing my Questions page, I am not thinking of some other 'GK' in some other world who took more care. I'm not the least bit interested in that. I'm interested in what I did, or, rather, failed to do.

Of your criticisms of possible worlds, I do need to respond to your argument that there are possible worlds where, owing to a fluke, physical laws are undiscoverable. OK, I agree. So what? In possible worlds (including the actual world) where physical laws are discovered, there is a satisfying explanation of how we are able to discover that those laws obtain. It is not an accident, so far as our beliefs are concerned, that those laws are true. That is sufficient for knowledge.

I am attracted by two alternative views of possible worlds:

(a) Possible worlds are sui generis. There distinction between what is and what might have been cannot be analysed further. Of course, that does not get us off the hook regarding the question of providing semantics for counterfactuals, even if we reject the idea that counterfactuals have truth conditions.

(b) Lewis is right and wrong about possible worlds. Right, because other possible worlds are as real as the actual world. Wrong, because the point of view from which the actual world is seen as merely another possible world is impossible for us, in the same way that it is impossible regard now as just another time or I as just another person (cf. Naive Metaphysics ch 18).

All the best,

Geoffrey