A.V. Ravishankar Sarma said:

What is the actual problem of counterfactuals in interpreting them in object language? In what sense
are counterfactuals relevant in technical areas like physics, computer science, AI?

============

Counterfactuals, or 'contrary-to-fact conditionals' as they are sometimes called, are one of the most
maddeningly difficult problems of philosophical logic. Take a simple statement like, 'If I hadn't been
working on these questions today, I would have been reading the book on Eastern Philosophy which
one of my students lent me.' Or, 'If I hadn't crashed my Ford Capri last April, I would still be driving it
now.' We make these kinds of statements all the time. We know when they are true, and when they
are false. (In fact, both these statements are false. Responding to the latest letters from my students
is higher up on my Do-list. And had I not crashed it, my Capri would certainly have failed its
compulsory Ministry of Transport inspection last July, and been sold or scrapped.)

But what makesa counterfactual true, when it is true? What kind of fact does a true counterfactual
statement convey?

Let's play a game of suppose. Suppose I didn't crash my Capri. There are lots of different situations
that I can imagine which are consistent with that supposition. 'If I hadn't crashed my Ford Capri, I
would have been driving a Lancia.' Or, 'If I hadn't crashed my Ford Capri, I would never have learned
to drive.' Intuitively, we know that these statements are absurd. Yet, one of the possible scenarios
where it is not the case where I crash my Capri, is the scenario where I never owned a Capri in the
first place. Another possible scenario is the one where I never owned a car in the first place.

Somehow, we are able to pick out the scenarios which are pertinentto the possibility that we are
imagining, and reject those that are not pertinent. The problem, as logicians have discovered to their
dismay, is that it is impossible to come up with a coherent set of rules for doing this.

One philosopher, David Lewis, has put forward the theory that the scenarios which are 'pertinent', in
this sense, are the ones existing in possible worldswhere the thing we are supposing is true, which
are most similarto the actual world. So, in the case of the supposition, 'I do not crash my Capri', the
most similar possible worlds are those in which I continue driving my Capri, rather than those in which
I never owned a Capri, or a car, in the first place.

It's a neat theory. The problem is that it doesn't altogether work. The idea of 'similarity', though it
works for some cases, gives the wrong results for others. One example that Davis Lewis himself
gives that causes difficulties for the similarity account is a statement like, 'If Oswald had not Killed
Kennedy, someone else would have.' This statement appears to come out true on Lewis' account,
simply because a possible world where Kennedy dies at the hands of another assassin is more
similar to the actual world than a possible world where Kennedy lives out his term of office. But
surely, one would only believe that someone else would have killed Kennedy if one thought (as many
in fact do) that Kennedy's assassination was the result of a conspiracy, rather than a single killer
acting alone. Philosophers are still waiting for a fully adequate logical analysis of counterfactuals.

Who else is waiting? Does anybody else care? Physicists, computer scientists, AI researchers etc. all
make use of counterfactual statements. To the extent that it is not fully understood how counterfactual
statements work,there will be a penumbra of unclarity whenever counterfactual statements are used.
I think it does matter, if scientists can't be sure what they are actually saying, or what their assertions
commit them to.

Geoffrey Klempner