Schaffer’s account of the hard cases

In document The chances of higher-level causation: an investigation into causal exclusion arguments (Page 173-179)

Chapter 3: A problem for counterfactual theories of causation

3.5 Schaffer’s account of the hard cases

Below, I reconstruct Schaffer’s account and I provide arguments to show that it is not a viable alternative to Maslen’s failed solution. When discussing the easy counterexamples to transitivity Schaffer (2005) follows the same general lines I described in (section 2.). According to him, even when contrasts are made explicit it is possible to formulate equivocal, in his words ‘shifty’, event nominals for the middle contrasts that can create the illusion of continuity. The logic should be clear from our easy example of the train from section 2., where using ‘to move not on T1’ as a contrast event description gave way to equivocation between ‘to move on T2’ and ‘getting derailed’. But Schaffer pushes this idea further, he argues that the source of confusion is the same concerning both the easy and hard examples, implicit ambiguity in the middle contrast descriptions, and he attempts to show that the identity of middle contrasts in a case like the assassin trainee (Ib) is as illusory as in the case of the runaway train (Ia).

This means that it should be true of hard cases like (Ib1*) that D1* in the first link and D2* in the second have different referents against all appearances. Schaffer’s analysis of the case would go as follows: ‘not to duck’ in the first causal statement of the scenario means that ‘the victim is upright when there is no command given’, whereas to make true the second statement of (Ib1*) D2* should mean something different like ‘the victim is upright when

there is a yelled command given’, so D1* ≠ D2*. The contrasts are different, because in the

first counterfactual link the contrast event on the effect side would take place in a world where there is no command given, whereas in the second link the contrast event on the cause side would take place in a world where the command is given already, because D2* □-› E* is true only in a world where the life of the victim is endangered by the shot, that is, where C occurs. If this is sound, we have a hidden contextual ambiguity in the middle contrasts and an explanation of why we don’t have a real causal chain in (Ib1*).

My contention is that this solution has unwanted consequences; it misidentifies some normal causal chains as counterexamples to transitivity. To see this one should first scrutinize Schaffer’s definition of events on grounds of which he justifies his solution. He, as all contrastivists, construes causation as a four-place relation where all the relata are events, to quote him “concrete, coarse-grained, worldbound occurrences” (Schaffer 2005:298). He explicitly commits himself to a view on the metaphysics of events that is, as he says “Davidsonian in spirit”. This seems to imply that the space-time region occupied by an event is among its identity conditions, but it is only a necessary not a sufficient condition. The other necessary condition we should consider is worldboundedness. Schaffer acknowledges some minor differences between the Lewisian construal and his view, the most important being that while Davidson is noncommittal with respect to the interworld relations of events, the contrastivist view takes events to be Lewisian individuals with counterpart theoretical

profiles. This means that events occurring in different possible worlds cannot be the same event and therefore events in the middle contrast positions are the same only if they take place in the same possible world68. If they don’t then even if they occupy the same space-

time region as their counterparts in other close worlds they should be considered as non- identical events. I am not sure whether these criteria together should be considered sufficient for event identity but for our present investigation it is enough to know that we are provided with these two independent necessary conditions.

Let’s check whether these conditions can deliver the right results for our hard transitivity counterexamples. If events were individuated only by the space-time regions they occupy then these conditional clauses (‘when there is no command given’, ‘when a yelled

command is given’) couldn’t play any role in differentiating the contrast events, because the

descriptions in the clauses refer to happenings in a space-time region that is distinct and distant from the core event. The yelling of the command and the absence of the command are events that cannot be part of the region that identifies ‘the victim is upright’ as a particular event, because the command is given (or not) outside and far away from the region that identifies it. Therefore, the space-time region criterion forces us to think that the two descriptions with different conditional clauses refer to the very same event. If the denoted event is the same event in both the first sentence of (Ib1*) and the second sentence of (Ib1*) and it makes both causal statements true, then (Ib1*) should be accepted as a proper causal chain.

68Trying to save the counterfactual theory other philosophers also subscribe to a counterpart theoretical

approach to events. Their idea is to build the context sensitivity of causal statements into the counterpart relation (e.g. McDonnell 2016).

However, on grounds of the worldboundedness criterion, it is plausible to interpret the conditional clauses in the middle contrast descriptions as expressing information about the world belonging of the events referred to, as expressing the fact that the two events are taking place in different possible worlds. Another way of thinking about this is to say that as the middle contrast events occur in different worlds, they inherit some traces, at least some slight influences form those different worlds and the clauses inform us about the world belonging of the middle contrast events by highlighting exactly these traces. When the victim stays upright in a world without a yelled command the victim has no soundwaves in his ears originating from the master assassin, whereas when he stays upright in a world with the command yelled, he has traces of that sound in his ears. The application of this second criterion suggests that ‘the victim stays upright when there is no command given’ and ‘the victim stays upright when a yelled command is given’ refer to different events and they do so because the corresponding events occur in different possible worlds. If one goes with this interpretation for the clauses, then Schaffer’s solution makes sense.

Now that Schaffer’s solution for hard cases is made plausible, we should check whether it really provides us with a precise enough tool, one that identifies apparent causal chains in the problems cases as cases of causal disconnectedness but preserves our intuitions concerning ordinary transitive cases. As I will show, it is not precise enough, because the difference in terms of world belonging between middle contrast events, the supposed illicit shift identified by Schaffer in short-circuits appears even in some straightforwardly transitive causal chains.

In scenario (Ic) we had a bull provoked by a stranger C that made it angry D which in turn made it to attack its owner E. As we saw, this is a perfect causal chain without any known problem, where C causes, provides proper control over E. As the solution suggested by

Schaffer aims to demarcate these normal cases from the non-causal, pseudo non-transitive cases, it should only apply to the short-circuit cases. But consider the following; let’s add a few details to scenario (Ic), imagine that C has an effect other than D, X, where X has no effect on the occurrence of D or E. Let’s call this version of the scenario (Ic1*). Its structure can be seen on Figure 3.5-1.

Adding such a detail is quite natural as most causes have multiple effects and this is all we need here, we need an alternative effect originating from C that leaves a trace on D. I would also risk the claim that such things occur quite frequently. Let’s say, event X results from C in the following way. By making his provoking moves the stranger also disturbs a bird that flies over the bull casting a shadow over it for a second. Now if we accept Schaffer’s criteria for event identity, this means that in the case of the first counterfactual the middle contrast D1* refers to the event where ‘the bull is tranquil when there is no shadow cast on

it by a bird’, but in the case of the second counterfactual D2* refers to ‘the bull is tranquil when there is a shadow cast on it by a bird’, as in this world C already took place causing X to

happen, casting a shadow on the back of the bull.

You can see on Figure 3.5-1 that the arrow originating from X avoids E, making it clear that this is not a short-circuit scenario. According to intuition there should be a straightforward transitive causal chain between C and E. However, in case we accept Schaffer’s criteria for event identity, we get a difference in the middle contrasts (D1* ≠ D2*) that amounts to an illicit shift, therefore, the chain between C and E is not a proper one.

To make things more transparent let us analyse scenario (Ic1*) in terms of the tests we should employ to test the truth of the two causal statements involved according to the contrastivist. First, we want to check whether C ‘the stranger is provoking the bull’ causes D

Figure 3.5-1 C



‘the bull is getting angry with a shadow cast on it’ is true. To do that, we should start from the actual world where C and D is in place, then fixing the past first, we move to the closest possible world where instead of C, we have contrast event C* ‘the bull is left alone, the stranger is just passing by’ and we check whether D1* ‘the bull is tranquil when there is no

shadow cast on it by a bird’ is in place as well. To do that, we have to fix the past and change

C to C*. The reason why in the first link of the chain D1* is ‘the bull is tranquil when there is

no shadow cast on it by a bird’ can be understood from how the scenario is constructed. When

C is in place the stranger disturbs the bird as well, not only the bull. So, C, as in the case of short-circuit scenarios, is a common cause of two events, D and X. But as C* does not excite the bird (X), in the counterfactual world where we should move to test the first causal link the shadow is not cast on the back of the tranquil bull either. Now, changing C to C* changes D to D1*, so the test gives the result that C causes D is true.

As a second step we should also check whether D ‘the bull is getting angry with a

shadow cast on it’ causes E ‘the bull is attacking its owner’ is true. To do that, we should start

from the actual world where D and E are in place, then fixing the past first, we move to the closest world where instead of D, we have contrast event D2* ‘the bull is tranquil with a

shadow cast on it’ and we check whether E* ‘the bull continues to sniff the flowers’ is in place

as well. The reason why D2* is ‘the bull is tranquil with a shadow cast on it’ can be understood from the details of the scenario. We know that the past of D contains C, the stranger had already disturbed the bird as well, not only the bull as C was a common cause of D and X. As a result, we change D to D2* in a context where event X remains in place. Now changing D to D2* changes E to E* so D causes E should be true.

Intuition tells us that C should also be the cause of E as it gives us perfect control over E. But Schaffer’s differential transitivity does not apply here, because on grounds of his criteria

D1* is a different event from D2* as it occurs in a slightly different world from the world where D2* does. So, my scenario seems to be a good counterexample to Schaffer’s solution for short-circuit scenarios. In other words, the application of Schaffer’s solution for short- circuits to a normal transitive scenario resulted in the unwanted consequence that we should analyse it as a deviant case. I take this to be a good enough reductio against the solution suggested by Schaffer.

To sum up, Schaffer relying on his special view concerning event identity tried to extend the contrastive solution for easy counterexamples to hard counterexamples. The idea was that in the case of both kinds of counterexamples non-transitivity is an illusion generated by the ambiguous nature of implicit contrast event nominals and by disambiguating them we can uncover the hidden crack in the causal chain. I have shown that in short-circuit cases this strategy fails, because the equivocation uncovered by Schaffer is such that it can be found in supposedly normal transitive cases as well.

In document The chances of higher-level causation: an investigation into causal exclusion arguments (Page 173-179)