How to handle the unexplained counterexamples?

In document The chances of higher-level causation: an investigation into causal exclusion arguments (Page 179-185)

Chapter 3: A problem for counterfactual theories of causation

3.6 How to handle the unexplained counterexamples?

If there is no better solution from the failure of the proposed solutions it follows that if we would like to maintain the contrastivist theory, we should accept the intransitivity of causation69. The problem with this stance is how to answer the questions put forward by

Maslen. Even though we can’t maintain the view that intransitivity is only an illusion created by sloppy descriptions, we still have to demarcate the non-transitive cases by some reliable

69 A similar idea is promoted by Hitchcock (2001) in the interventionist theory of causation which is quite close

to the contrastivist. According to Woodward (2005) the interventionist theory is a special case of the contrastivist theory. The main difference between their view on transitivity and Maslen’s is that these authors accept all kind of counterexamples to transitivity as valid, not only the short-circuit cases.

means. This is the only way to preserve the explanatory power derived from transitivity where it is required.

Some people working on different theories of causation suggest that the only reason we are still struggling with the hard counterexamples is that we mistakenly take negative causation to be real causation (e.g.: Moore 2009). Short-circuits involve preventions of other events in the second link and prevention is a case of negative causation, but if negative causal talk is not ontologically serious then these counterexamples are only relevant for causal explanation, not for causation itself. Most philosophers working on process theories or power theories of causation agree with this image. (Dowe 2001, 2004 Mumford and Anjum 2011). Some argued for the same idea even in the context of counterfactual theories (Beebee 2004) This solution has its advantages, but most supporters of the contrastivist framework and many believers of counterfactual accounts in general are explicitly committed to ontologically serious negative causation for good enough reasons (see Schaffer 2004, 2005; Lewis 2004; Menzies 2006), therefore pursuing this line of thought requires a thorough discussion of the issue which exceeds the limits of this discussion. I will only consider answers that are directly available for the contrastivist.

As there are no known hard problem cases other than the short-circuit scenarios the demarcation might be done quite trivially. One might say that whenever we have a proper short-circuit scenario, we have a case of non-transitivity at hand, and there are no other cases. The question is how to identify short-circuits properly? The solution I will argue for originates from a sentence-long footnote in Maslen’s paper. She mentioned the idea as a possible alternative to her backwards counterfactual criterion. I believe it is the only working solution. The statement below should be true in cases of proper transitive chains and it should be false

when and only when it comes to short-circuit type scenarios (adapted from Maslen 2004:357, footnote 35):

(C* & D*) □-› E*

Think about a cascade of dominos as a straightforward case of a transitive causal chain. It is true that if neither the first nor the second domino had fallen then the third domino wouldn’t have fallen either. This works in the case of the angry bull (Ib) as well. If the stranger had walked along and the bull had continued grazing than the bull wouldn’t have attacked its owner. However, the criterion is false for short-circuit scenarios like the assassin trainee. It is false to say that if the master assassin had stayed silent and the victim had stayed upright then the victim would have died. This sounds promising. It is even surprising that Maslen chose to work out the more problematic backwards counterfactual criterion in her paper. This suggestion seems to work and it is devoid of the theoretical difficulties of her main proposal. The only problem one can raise is that the criterion is unmotivated; we get no explanation of why it is capable of doing the job of demarcation.

To give an interesting explanation one should reformulate what the criterion has to say. In short-circuit scenarios the occurrence of the first event of the causal chain in focus (C- D-E) and the alternative chain of events the first event initiates (e.g. the trainee pulling the trigger in (Ib1*)) is a precondition for the truth of the second causal statement, D causes E., in the (C-D-E) chain. Or to put it differently C serves a dual role here, by transitivity it is the cause of E and it is also among the background conditions of the causal relation between D and E70. The counterfactual if C hadn’t occurred then D rather than D* wouldn’t have caused

70 The occurrence of effects always depends on a number of causal conditions even though in everyday practice

we select one as the cause of the effect. Most counterfactual accounts are indifferent with respect to causal selection they don’t differentiate between causes and background conditions. From the perspective of those

E rather than E* holds true for these cases. In (Ib1*) if there is no loud command given then there is no shooting, but if there is no shooting then ‘ducking’ (D) rather than ‘being upright’ (D*) does not make a difference to the occurrence of the survival of the victim (E).

In normal transitive scenarios like domino cascades or in cases like (Ic1*) this is not true. The bull getting angry (D) rather than remaining calm (D*) does make a difference to whether it attacks its owner or not no matter the stranger was there to excite it (C) or not (C*). For the fall of the third domino in a queue of dominos the presence of Earth’s gravity is a background condition and we would probably say it is caused by the fall of the second domino. The fall of the third domino counterfactually depends on both of these causal conditions and there are other events on which it does not depend, e.g. the lighting conditions of the space where the third domino falls. Interestingly, it is true about the fall of first domino as well that it is not among the background conditions of the fall of the third domino. The fall of the second domino can cause the fall of third in the absence of the fall of the first domino. At the same time, one could argue that the first domino can cause the third to fall via causing the second to fall. But even if it does, the capacity of the second domino to cause the fall of the third is not dependent on the fall of the first domino. As we saw, this is not true of short- circuit cases and that is what leads to the existence of non-transitive causal chains.

Unfortunately, I can’t prove that the identified structural feature only appears in short-circuits, but on grounds of our limited induction base, what we know about examples of transitivity and counterexamples to it, the criterion that the occurrence of an earlier event

frameworks any event counts as a causal condition of an effect if the occurrence of the effect depends on that event counterfactually and, naturally, not all past events are among the causal conditions of an outcome. Here, I rely on the everyday distinction between the cause and other causal conditions and will call the latter background conditions because the distinction is helpful in highlighting the dual role the first event plays in short-circuit scenarios.

in a causal chain is a precondition for the truth of a later causal statement in the same chain seems to do the job of demarcation, it identifies short-circuits and only them. At present I don’t see any good reason to think that the above structural characterisation is not good enough to identify all problem cases relevant for transitivity, so I suggest it as a temporary solution for the problem of demarcation.

However, there is an important issue one should consider, short-circuits can come in many shapes and colours they might involve more links, more causal lines cancelling each- other out or they might have other less predictable variations. If they share those basic structural features I have highlighted above, we might be able to identify all of them. Proving this properly is a big enough project in itself and I won’t pursue to do that here. And there is even more to this problem, it is well known from the literature on actual causation that the sheer number of possible causal structures is beyond human control (see: Glymour et al. 2010) and if it is, we simply can’t be sure that there are no further counterexamples beyond those we know about at present. So, the tentative solution I suggested is only a solution until we encounter some further tricky counterexamples, but at least till that time it could be a useful suggestion.

3.7

Concluding remarks

This chapter has shown that contrastivists cannot demarcate or explain away short-circuit type counterexamples to the transitivity of causation. Maslen was unsuccessful in formulating a working demarcation condition for non-transitive cases. Schaffer tried to extend the solution developed for easy counterexamples to cover the hard cases and explain them away, but it turned out that if one treats the middle contrast events of short-circuits only as close counterparts in different worlds to resolve the issue unacceptable consequences follow. The counterpart events story led us to the problem that some straightforward causal chains do not conform to the suggested criterion. From all this I drew the conclusion that there are no good enough methods for the contrastivist to explain away the hard counterexamples, but I have suggested that it is possible to demarcate them based on certain structural features and thereby to save those theoretically important explanations that rely on transitivity.

So, it seems that counterfactual theories of causation have their difficulties as well, even if they perform better on a lot of tasks than productive theories of causation, but the research program built around the basic counterfactual notion of causation still seems to be a progressive one.

In document The chances of higher-level causation: an investigation into causal exclusion arguments (Page 179-185)