The basic contrastive account of causation

In document The chances of higher-level causation: an investigation into causal exclusion arguments (Page 164-167)

Chapter 3: A problem for counterfactual theories of causation

3.3 The basic contrastive account of causation

Below I will show how the framework is applied to solve the easy counterexamples and, in the sections, following that I will discuss and criticize solutions to hard counterexamples. Contrastivist solutions to the problems of the classic counterfactual theory presuppose that the surface form of everyday causal statements is misleading. Even though on the surface causation seems to be a two-place relation, causal statements tacitly express a four-place relation. Causes and effects have contrast events attached to them. The basic scheme is as follows:

C rather than C1* causes E rather than E1*

Contrastivists maintain that by making contrasts explicit, in other words by making the worldly context of the causal statement explicit, many problems of the old counterfactual theory can be solved, the problems of transitivity, selection64, negative causation, the graining

of events and so on (see Schaffer 2005). The origin of these problems is identified in ambiguities that vanish if the proper contrast events, the context relevant for the statement is taken into consideration. It is easy to make any causal statement true or false by arbitrarily choosing contrasts for the surface statements, but it is also straightforward that in everyday practice we rely on a tacit system of rules to decide on the admissibility of contrasts relative to a particular context. Recently Reiss (2013a, 2013b) developed interesting ideas to account for this practice. Even though it is a received view in the relevant literature that the choice of contrasts is determined by the context a satisfactory account of the mechanism by which context determines the meaning of causal statements, the origin of the contrast classes

64 Selection is the practice of choosing one cause as the cause among the many causal factors contributing to an

remained elusive (see Schaffer 2013). This challenge, however, is not particularly relevant to the subject of this chapter and will not be considered here. Below, I presuppose that the basic contrastivist idea is sound enough and constrain the discussion to the solutions provided for the problem of transitivity.

In all of the proposed contrastive theories it is accepted that to test the truth of a causal statement one should test the truth of the counterfactual involving the contrasts events associated with the explicit cause and effect event (see: Maslen 2004; Schaffer 2005). It is the truth value of the counterfactual C1* □-› E1* relative to the actual world where both C and E occurs that decides the truth value of the causal statement ‘C causes E’ instead of C □-› E suggested by the surface form. These truth conditions were introduced to make the truth evaluation of causal statements relative to a specific context; the contextual features relevant for the truth evaluation are expressed by the contrasts.

The transitivity of causation should also be reformulated for this quaternary approach. It would mean that for all (x, x*,y, y*,z, z*) if x rather than x* causes y rather than y* and y rather than y* causes z rather than z* then x rather than x* causes z rather than z*. To demonstrate the power of the contrastive idea, it is useful to consider how the contrastive method works in the case of the runaway train (Ia). As we saw above, the contrastivist proposes that for a causal chain (meaning at least a chain of two causal links) to be a real causal chain it has to be continuous or linked up at the middle event (that is, the effect event in the first link and the cause event in the second link) not only on the explicit side but also on the implicit, contrast side. This is called ‘differential transitivity’ by Schaffer (2005:308). The contrasts can’t simply be negations of the explicit side statements; the context provided by the particular causal scenario helps us and requires us to make the contrasts more precise

than that, but the contrast should be an event that is compatible with the negation of the explicit side event. Let us first analyse (Ia) using negations as default contrasts (Ia1*):

The position of the pointer on T1 (C) rather than not on T1 (C*) causes the train to move towards the station on line T1 (D) rather than to move not on T1 (D1*).

The train moving towards the station on line T1 (D) rather than moving not on T1 (D2*) causes the train to run into the station building (E) rather than not to run into the station building (E*).

The position of the pointer on T1 (C) rather than not on T1 (C*) causes the train to run into the station building (E) rather than not to run into the station building (E*). On grounds of the above formulation one can say that the last causal statement is clearly false, but both 1. and 2. seem fine and if this is all right then the counterexample holds. Fortunately, taking a closer look at the contrasts, one can show that 2. is false as well. Because of the apparent identity of the descriptions in D1* and D2*, we might come to believe that the scenario is a case of transitivity. But ‘to move not on T1’ is equivocal. In the case of D1* the context of the scenario suggests that ‘not on T1’ has to be ‘to move on line T2’ but if one gives D2* the same interpretation it makes statement 2. false. To make 2. true ‘not on T1’ should refer to something like a derailment, where the train is not moving neither following track T1 nor following T2. But according to what we know about the situation no derailing is involved in the context of the scenario and furthermore if D1* would refer to a different event than D2* then the two statements would belong to different contexts, so we would have no reason to think that they belong to a single chain of events. Here we should remember that a valid transitive causal chain requires two properly connected, true causal statements. By disambiguating the middle contrast descriptions we might get two true causal statements, but in that case there is no continuity at the middle contrast. It is also clear that if D1* and

D2* are identical then we have no real causal chain either, because when they are (D1*=D2*=‘to move on line T2’, or D1*=D2*=‘not to move on either line, derail’) one of the two causal statements becomes false. The conclusion from all this is that in scenario (Ia1*) it would be wrong to accept 1. and 2. together as a proper causal chain as either 2. is not a valid causal statement or 1. and 2. aren't linked up, so it is no surprise that statement 3. is false. If there is no real causal chain involved in a scenario then the question of transitivity is irrelevant, so the counterexample is explained away.

Maslen and Schaffer agree that either a causal chain is properly linked up at the middle contrast and consists of valid causal statements or the presumed causal chain is not a causal chain at all. To have a proper causal chain the following statements should all be true about the contrasts: C*□-›D1*, D2*□-›E*, D1*=D2*. To sum up, the basic contrastivist understanding of the counterexamples is this: in cases of perceived non-transitivity we are misled by the ambiguities of the surface form of everyday causal statements. Intransitivity is an illusion created by sloppy descriptions.

We should now turn to the hard cases. People who tried to systematize the contrastive view recognized that these cases are different, but they disagreed on the required solution. Below I discuss the two known suggestions to argue that they require supplementation.

In document The chances of higher-level causation: an investigation into causal exclusion arguments (Page 164-167)