Cross-posted from http://mleseminar.wordpress.com/
...
Boris Kment - Counterfactuals and Explanation
A really interesting paper this week - it can be found here, and the presentation is here.
Kment's main proposal is that match of matters of particular fact
should be relevant to closeness of two worlds for the purposes of
evaluating counterfactuals if and only if the matters of fact have the
same explanation in both worlds. Secondarily, he proposes that we
should allow for laws to have exceptions, and hence that all worlds
which share the same laws as ours should be closer to actuality than
any world with different laws.
We quite liked the main proposal, but worried about the
individuation of explanations it relies upon. What are the conditions
for two events to have the same explanation? For example, consider the
counterfactual 'if I had tossed the coin five minutes earlier, it would
still have come up heads'. This seems false, but perhaps Kment can
account for this falsity by saying that the coin's coming up heads in
the various A-worlds would have a different explanation from its
explanation in the actual world, because it would have been caused by a
different event of tossing.
However, what about 'if I had tossed the coin one nanosecond
earlier, it would still have come up heads'? Here we were much more
inclined to take the counterfactual as true. Perhaps this difference
goes along which goes along with the intuition that the actual tossing
and the 1-nano-second earlier tossing count as the same event (or as
counterparts according to some very natural counterpart relation),
while the actual tossing and the 5-minute-earlier tossing count as
different events. But if this is the line Kment would want to take,
we'd need to hear more about how it is to work.
Finessing the individuation critera for explanations might also
afford a solution to the problem case (25) which Kment mentions
inconclusively. If the explanation for the lottery's having the result
it did does not include that phone A was used to make the call, but
just includes that some phone of such-and-such qualitative character
was used to make a call, then we would get the right result that even
if phone B had been used, the result of the lottery would have been the
same. This requires that the explanation of the lottery's result should
only include qualitative features of certain early-enough explanatory
factors, rather than the whole fully-detailed causal story. That is,
explanations should comprise roughly the minimal information required
to determine their explanandum.
This solution involves dropping the transitivity of explanation
which Kment explicitly assumes - because it is plausible that a call
being made explains the outcome of the lottery, and that the use of
phone A explains that a call was made. However, perhaps dropping
transitivity of explanation is any case desirable. Consider the
well-known counterexample to transitivity of causation - the boulder's
rolling down the mountain is the cause of the hiker's ducking, and the
ducking is the cause of his survival, but the boulder's rolling is not
the cause of survival. The same counterexample seems to work against
transitivity of explanation - the rolling explains the ducking, and the
ducking explains survival, but the rolling does not explain survival.
Another issue we thought about was the degree to which a Humean
could adopt the notion of laws as having exceptions. Clearly it's
incompatible with Lewis' own theory of laws, according to which the
laws are those true universal generalizations which provide the best
balance of simplicity and strength, but perhaps (as Antony suggested) a
Humean view which took laws to be more like habitual statements would
work. Habituals tolerate exceptions, but they still explain their
instances.
Maria had a potential objection to this approach for the Humean (and
to any view according to which there are restrictions on how many
exceptions are possible before the laws have to be different) - suppose
the number of exceptions in a world are right on the borderline for
it's having some particular laws. Then the extra small exception needed
to accommodate some antecedent would involve consideration of an
A-world with too many exceptions to have the same laws as the original
world. Then the A-world which, intuitively, is the right one for
evaluating the counterfactual would not come out as closest according
to Kment's criteria. So it looks like the view might only in fact be
compatible with strong 'immanent' views of laws where any arbitrary
number of exceptions are possible while the laws remain the same.
One final thought; it would be possible to hold that exceptions are
possible to all special-scientific laws, but not to the fundamental
laws, if such there be. This seems to fit well with usage: we talk of
the 'laws' of statistical mechanics, even though they only hold with
high probability, but we are much less willing to admit that the laws
of fundamental physics might have exceptions. Someone who took this
view of laws could carry over everything Kment says about ordinary
counterfactuals, though might have to say something a little more
counterintuitive about counterfactuals concerning fundamental physics
(perhaps in a deterministic world we would have to count as true 'if
this electron had been over here and not over there, the matter
distribution at the big bang would have been different'). However, this
consequence might be ameliorated by the indeterminacy of fundamental
physics.