Posts on: Causation
Gallow (2023) spells out an interventionist theory of counterfactuals that promises to preserve two apparently incompatible intuitions.
Suppose the laws of nature are deterministic. What would have happened if you had chosen some act that you didn't actually choose? The two apparently incompatible intuitions are:
(A1) Had you chosen differently, no law of nature would have been violated.
(A2) Had you chosen differently, the initial conditions of the universe would not have been changed.
Rejecting one of these intuitions is widely thought to spell trouble for Causal Decision Theory. Gallow argues that they can both be respected. I'll explain how. Then I'll explain why I'm not convinced.
In my recent post on Interventionist Decision Theory, I suggested that causal interventionists
should follow Stern and move from a Jeffrey-type definition
of expected utility to a Savage-Lewis-Skyrms type definition. In that case, I also suggested that they could avoid various problems arising from the concept of an intervention by construing the agent's
actions as ordinary events. In conversation,
Reuben Stern convinced me that things are not so easy.
Brian argues that our intuitions about whether an action C causes somebody's continued survival is linked to the applicability of causative notions like "opening", "closing", "protecting", "threatening": if C inadvertently causes the survivor to be threatened but at the same time protects him from the threat, we are more inclined to count C as causing the survival than if C threatens the surviver but at the same time inadvertently causes him to be protected.
This argument looks a lot better than it is:
Suppose some physical event E is causally necessitated by a certain distribution of physical properties P. Then if P occurs, E is bound to occur as well, no matter what else is the case. In particular, whether or not some non-physical event M also occurs before E will make no difference to E's occurrence. (Perhaps M nevertheless causes E, if E is overdetermined, or perhaps M is causally relevant in some even weaker sense, but at any rate M does not make a difference for E.)
To see the problem with this argument, consider a deterministic world where the occurrence of any event E at time t0 is causally necessitated by the state of the world at t-2 (before t): it obviously does not follow that the state of that world at t-1 makes no difference to E's occurrence.
About half a minute ago, I've poured tea into this cup. In a few seconds, I will take a
sip. What if I had taken a sip a minute earlier? I wouldn't have taken
a sip of tea from an empty cup: that is impossible. So there would
have been tea in the cup a minute ago. How did it get there? Maybe
I would have poured it in earlier. Or maybe it would have tunnelled
directly from the pot into the cup. Or maybe the tea would have
just materialized out of thin air. Some of these counterfactuals
do not sound very plausible, but let's assume that for the kind of
counterfactuals relevant to causation, they are all equally good so
that there is no fact of the matter about how the tea got into the cup
at the closest world where I take the sip a minute earlier: it does
so differently at different worlds that are equally close. (See Lewis,
"Counterfactual Dependence and Time's Arrow" for the standards of
evaluating such counterfactuals, and "Are we free to break the
laws?" for the indeterminacy of divergence miracles.)
Merlin is bound to disappear at noon, taking with him all physical
traces of his existence. Shortly before his magic disappearance, he
casts a spell. As a result, at noon on the following day, the prince
turns into a frog.1
In virtue of what does the spell cause the metamorphosis? For
instance, it is not at all clear that by Lewis's standards of
similarity, some world containing neither spell nor metamorphosis is
more similar to actuality than any world not containing the spell but
containing the metamorphosis. The problem is that the only trace left
by the spell, after Merlin's magic disappearance, is the
metamorphosis itself:
This appears to be a problem for Lewis' theories of causation:
Let A,B,C,D be any events such that B depends counterfactually on A, and D
on C. Now consider the conjunction (fusion) B+C of B and C. If A had not
occurred, B+C would not have occurred. For then B would not have occurred,
and presumably B+C can't happen without B. And if B+C had not occurred, C
would not have occured either, so (unless the absence of B has some
surprising effects on D), D would not have occurred. Hence there is a
chain of counterfactual dependence between A and D. But since A,B,C,D were
arbitrary, this means that every cause causes every effect.
If you're asked to explain how your preferred theory of everything -- that is, your brand of physicalism -- can accomodate some entity X, the first thing to try is the Canberra Plan. It goes as follows: First, collect features that could be said to characterise X. If you're lazy, simply collect everything the folk says about X. Next, say that since these features comprise the essence of X, whatever physical entity has (more or less exactly) those features is X. Finally, explain that of course there is such a physical entity, since otherwise statements about X wouldn't be true.
An old puzzle: The average mother has 3.4 children. Yet the average
mother does not exist. So how can she have children? An old solution: She
doesn't. "The average mother has 3.4 children" is to be understood as
"the number of children divided by the number of mothers is 3.4". So
"average mother" is not a genuine predicate, but rather a meaningless part of
numerical predicates like "the average mother has ... children".
If this solution is correct, it is meaningless to say that average
mothers exist, that some of them influence others, and that all of them
are distinct. Which indeed it is.