[ Pobierz całość w formacie PDF ]

can, with some adjustment and simplification for the purposes of discussion, deal
with a range of problems (see Noordhof 1999 for original presentation of the
theory). Specifically, I will be concerned with the issue of whether the semantics
of counterfactuals can be characterized independently of causation (raised by
Dorothy Edgington, this volume), the proper way to deal with the nontransitivity
of causation (raised by Michael McDermott 1995 and Murali Ramachandran, this
volume), and a collection of counterexamples to the idea that causation involves, at
its heart, chance-raising (discussed in this volume by Helen Beebee; Phil Dowe;
Doug Ehring; Chris Hitchcock and Michael Tooley, and by Jonathan Schaffer
(2000a, 2000b)). Obviously, in defending my own counterfactual theory, I am also
implicitly arguing that counterfactual approaches to causation in general have
the resources to capture its important features. The ambiguity in the title thus
accurately reflects the content of the present paper.
1 My theory (or at least part of it)
My central idea is that causes are events which, if they occurred independently of
their competitors, would make the chance of their effects very much greater than,
at the limit, their general background chance at the time at which the effects
occurred via an actually complete causal chain. Hence I do not take causal claims
to be essentially contrastive: true or false relative to whichever alternative
scenario is had in mind (see for example Hitchcock (forthcoming)). I appealed to
counterfactuals and various other notions in order to capture this idea. A simplifi-
cation of my theory runs as follows:
For any actual, distinct events e1 and e2, e1 causes e2 if and only if there is a
(possibly empty) set of possible events � such that
(I) e2 is probabilistically �-dependent on e1, and,
A counterfactual theory of causation 189
(II) every event upon which e2 probabilistically �-depends is an actual event,
(III) e2 occurs at one of the times for which p(e2 at t) e" x>> y.1
I define probabilistic �-dependence in the following way:
e2 probabilistically �-depends upon e1 if and only if:
(1) If e1 were to occur without any of the events in �, then for some time t, it
would be the case that, just before t, p(e2 at t) generally around x,
(2) If neither e1 nor any of the events in � were to occur, then for any time t, it
would be the case that, just before t, p(e2 at t) generally around y,
(3) x >> y.
Let me offer a few preliminary comments in the way of explanation. Other
features of the proposal will become clearer when I turn to some of the problems
that have been raised for approaches like mine.
Talk of �-dependence is a mechanism by which to take away competitor
possible causes, for example, in cases of preemption or over-determination. When
there is a preempted or back-up chain, it need not be true that what we might intu-
itively count as a cause would be necessary in the circumstances, or significantly
raise the chance of an effect. However, it would still be true if events in the
preempted chain did not occur (that is they were put in the �-set). The definition of
probabilistic �-dependence defines a notion of chance-raising conditional upon
the events in the �-set being absent. The time of assessing the chance of the puta-
tive effect is just before the occurrence of the effect. If we put an event of the
preempting causal chain into the �-set, then the preempted chain can run to
completion. That often means that there will be an event, which didn t actually
occur, occurring in these changed circumstances. We don t want to conclude that
the preempted chain contains causes of the putative effect. After all, it was
preempted. Clause (II) rules this out by insisting that there should be no non-actual
events upon which the putative effect probabilistically �-depends. We just saw
that in the case of the preempted chain this may well not be the case. Of course, the
preempted chain may not be filled in. But in that case, it will not be true that the
chance of the putative effect assessed just before its occurrence is raised by the
occurrence of the putative, in fact preempted, cause.  x >>y should be read as x is
proportionately very much greater than y. This does not mean that the probability
of x should be high. Clause (III) and assessing the chance of e2 occurring at a time t
ensures that we are not just considering whether the chance of the putative effect,
e2, is raised but more importantly whether it is raised by e1 at the time that e2 actu-
ally occurred (for details and further discussion see Noordhof 1999: 108 14).
I appeal to the idea of chance-raising because I think it is plausible that there are
indeterministic causes. Some have denied that appeal to chance is necessary (such as
Ramachandran (1997), Barker (this volume: 120 4, 132 4)). Simple counterfactual
dependence is all that is required. I remain unconvinced (see Noordhof 1998a: 459 60).
190 Paul Noordhof
It seems to me that those who deny that appeal to chance is necessary face one challenge
and have one unargued commitment. The challenge is to explain an asymmetry. It is
alleged that, in every putative case of indeterminism, even though there is not a sufficient
cause of a certain effect, there will be conditions which, together with the absence of the
putative cause, will be sufficient for the effect to have no chance of occurring.2 Why [ Pobierz całość w formacie PDF ]