Causation Isn t Contrastive It s Contextual

Transcription

1 Causation Isn t Contrastive It s Contextual

2 Outline Over 35 years after David Lewis pronounced, It remains to be seen whether any regularity analysis can succeed... without piling on the epicycles it looks as though his own counterfactual account of causation is meeting the same fate: no end to epicycles and counterexamples The latest move in this game has been to declare causation to be contrastive relative to explicit contrast events In this paper I want to show that contrasting solves none but the simplest problems, and for a reason Other cases are not dealt with unless the theory is amended by an account of admissibility (of contrast events)

3 Outline I will show that specific kinds of contextual features determine admissibility I will then argue that once context is included in the semantics of causal claims, there is no need for explicit contrasts Moreover, a further set of cases (involving the timing of the events) demonstrate that the context determines more than just whether C causes E : it also determines which of a number of causal notions ( causes / prevents / affects the manner of ) is the correct one to be used First, let us look at the simple counterfactual account though

4 The counterfactual account Roughly speaking, C causes E if Had C not been, E would not have been is true (sufficient condition) According to this account, causation has these characteristics: it is a relation that holds between (token) events; this relation is one of counterfactual dependence, which is understood in terms of similarity among possible worlds; it is transitive; it is non-perspectival (or absolute).

5 ... and its discontents There are counterexamples related to each of these characteristics (for instance): 1. (Counterfactual dependence redundant causation) 2. A relation between events omissions/preventions, emphasis, fragile events 3. Similarity among possible worlds many-valued variables 4. Transitivity Non-transitive causal relations, aspect causation 5. Non-perspectivality selection Understanding causation as contrastive is supposed to help with all of these except (1), naturally

6 The contrastive theory C rather than C* causes E rather than E* if C, E are actual, distinct events C*, E* are alternative possible events E* counterfactually depends on C*

7 Problems contrasting can handle: Emphasis Leonike s bike (Dretske 1977). Leonike is interested in purchasing either a bike or a scooter she has seen in the window of a nearby store. She lacks the money to buy either, so one night she breaks into the store and steals the bike, the easier vehicle to take. Later that night she is arrested by the police. Did Leonike s stealing the bike cause her arrest? Yes: Leonike s arrest depends on her stealing the bike No: Leonike s arrest does not depend on her stealing the bike Solution: specify C = Leonike s stealing the bike; E = her arrest; E 1 = Leonike remains free Alternative 1: C 1 = Leonike buys the bike Alternative 2: C 2 = Leonike steals the scooter

8 Problems contrasting can handle: Many-valued variables Railroad assassin. Assassin wants to derail an express train by changing a railroad switch and thereby diverting it onto a dead track. Unbeknownst to assassin, there is a third setting, local, to which he accidentally sets the switch. The train arrives safely. Did the assassin s action cause the train to arrive safely? No: What he did had no influence on the safe arrival of the train. Yes: The assassin failed. Had he not failed, the train would have derailed. His failure caused the safe arrival. This is a problem of the similar-worlds account because that yields one clear answer: The assassin s failure caused the safe arrival

9 Lesson 1 Claim: Contrasting helps with paradoxes of the counterfactual account But only in cases where the intuitive causal judgement is ambiguous Analysis (given by contrastivists): causation is contrastive; the intuitive judgement is ambiguous because there are several equally legitimate implicit alternative events; contrasting explicitly resolves this by disambiguation

10 Lesson 1 Claim: Contrasting helps with paradoxes of the counterfactual account But only in cases where the intuitive causal judgement is ambiguous Analysis (given by contrastivists): causation is contrastive; the intuitive judgement is ambiguous because there are several equally legitimate implicit alternative events; contrasting explicitly resolves this by disambiguation

11 Contexts? Schaffer 2005 argues that the context determines the contrasts He offers a two-stage theory of causation: counterfactual dependencies determine the truth value of causal claims pragmatics (e.g., presuppositions) determine what causal claims are appropriate to make

12 Hard cases: Omissions and Preventions The gardener. Billy and Suzy have grown up. Suzy is now a famous movie actress and owns a mansion. Billy hasn t been so lucky and would be job and homeless if it hadn t been for Suzy giving him bread and a roof in return for him taking care of her garden. But Billy grows bitter over their unequal relationship and one time, when Suzy is away from home to shoot a film, leaves her mansion. When Suzy returns, she finds her favourite plants dead. True: The garderner s napping rather than watering the plants caused their death False: The queen s doing queenly things rather than watering the plants caused their death True: Had the gardener watered the plants they wouldn t have died. True: Had the queen watered the plants they wouldn t have died.

13 Hard cases: Omissions and Preventions A related case... The spurious breakfast. I had breakfast this morning. Now I m giving a talk. Did my having breakfast cause my giving a talk? Intuitively, no. However, here are some alternative events that make it a cause: C 1 - my eating of the poisoned dish; C 2 - my perishing in a meteorite shower Call this the problem of spurious preventers. (This may be less a problem for Lewis than it is for the contrastive account)

14 Hard cases: Modifiers Modifiers are events that change a property of an event without causing the event: The blue fire (Ehring 1987). Davidson s adding copper chloride caused the fire to be blue but it didn t cause the fire. This can be a problem for transitivity: if the fire (for example) causes someone s injury The contrastive analysis has difficulty with cases such as this because, although there exists a set of contrasts that identify the causal judgement correctly, there can also exist an alternative set of contrasts that gets the judgement wrong: Davidson s adding copper chloride rather than calcium chloride caused the fire to be blue rather than orange (true) BUT: Davidson s adding copper chloride rather than water caused the fire (false)

15 Difference from first three cases The second set of cases differs from the first set in that (1) the intuitive judgement is clearcut; (2) while there is indeed counterfactual dependence relative to some contrast events, the corresponding causal judgement is false They therefore also show that Schaffer is mistaken: it s not just inappropriate to say that the queen caused the wilting of the plants, it s false

16 Specifying alternative events The problems of proliferation of causes, spurious preventers and modifiers demonstrate that we have to specify very carefully what is admissible as an event alternative Clearly, C* and E* cannot be identical to C and E, respectively: they have to be genuine alternatives Northcott: C/C* and E/E* have to be nomologically incompatible ; i.e., there is no world in which the same laws are true as in our world and the alternatives co-occur But this is too weak and too strong at the same time: my eating of the poisoned dish is compatible with my having breakfast but can be admissible in the right circumstances; being struck by meteorites is incompatible with my having breakfast but inadmissble

17 Contrasts in context It seems safe to say that contextual features that determine admissibility My question now: what are the kinds of contextual features that determine alternatives? a) Situational features b) Analyst s features

18 Situational Features Captain and trainee assassin (Hitchcock 2003; Northcott 2008). Captain yells fire, trainee fires. Upon hearing the command, victim ducks. The bullet misses him and victim survives unscathed. Did the command cause victim s survival? Contrasting allows us to disambiguate: No: C 1 - captain gives no command Yes: C 2 - captain gives silent command C 1 is admissible if captain intends to kill victim, there is background noise etc. C 2 is admissible if captain is under orders but wants to save victim s life

19 Analyst s Features

20 Analyst s Features

21 Analyst s Features

22 Analyst s Features II Hart and Honoré provide a contrastive account of causation in which causation is relative to some normal state of affairs This normal state of affairs is often given by lawful action of the agents involved The Empire Jamaica (Hart and Honoré 1985; cf. Schaffer 2005). The owners of the ship sent it to sea without properly licensed officers. The ship was later involved in a collision. The pilot, though unlicensed, was generally competent but napping at the time of the collision. Did the owners action cause the collision? Here the relevant contrast event is given by the owners lawful action, which may or may not have been in their range of options; in particular, to determine causation, it is unnecessary for the defendant to have known the law (ignorantia iuris non excusat)

23 Contrasts in Context Call an event objectively possible when it is expectable from the point of view of the agents involved in the situation (on the basis of their beliefs, desires and opportunities) or the language user making the causal claim (and their presuppositions, normative commitments and rules of inference) Is causation then 5-place? No! Once contextual, contrasting is redundant Contrastive causal claims are clumsy The surface grammar of causal claims is not contrastive

24 Contextual Causation The context is thus: X = {B, D, O, P, N, L} Situational beliefs (B), desires (D) and opportunities (O) Analysts presuppositions (P), normative commitments (N), principles of inference (L) Contextual Causation. In X, C causes E if C, E are actual events; C*, E* are alternative, objectively possible events in X; and E* counterfactually depends on C*.

25 Asymmetry of delayers and hasteners Delayers and hasteners are special kinds of modifiers that affect only the time of the event Typically, hasteners are called causes whereas delayers are not: assassins kill and doctors save lives Asymmetry is typically understood to lie in some universal metaphysical difference: Hastening closes off options whereas delaying doesn t (Schaffer) Preventing only by causing is causing whereas causing only by preventing isn t (P. Mackie)

26 Hasteners and Delayers in Contrast Constrasting cannot mark out that difference: True: Assassin shot rather than kissed victim, which caused him to die rather than blush (i.e., live) False: Assassin shot victim rather than put the gun down, which caused him to die at t 0 rather than at t 1 > t 0 (while it would be correct to say: Had the assassin not shot, victim would have died later) True: Doctor performed the tracheal intubation flawlessly rather than improperly, which prevented the patient s death False: Doctor performed the tracheal intubation flawlessly rather than improperly, which caused the patient to die at t 0 rather than at t -1 < t 0 (while the counterfactual statement is true) That is, the contrastive treatment is exactly parallel while the judgement is reversed ( causes versus prevents )

27 Hasteners But things get worse: though hasteners are usually causes, this isn t always the case

28 Hasteners But things get worse: though hasteners are usually causes, this isn t always the case Bob Marley. I shot the sheriff and thereby caused his death at t: C* - I put the gun away; E* - the sheriff died (by gunshot) at a time later than t

29 Hasteners But things get worse: though hasteners are usually causes, this isn t always the case The disinterested doctor. A doctor induces labour in a becoming mother and thereby affects the timing of the birth without causing it

30 Hasteners But things get worse: though hasteners are usually causes, this isn t always the case The considerate doctor. A physician induces labour and thereby prevents the child from having to fight in the Vietnam war (because only men born between July 1 and December 31 are drafted)

31 Hasteners But things get worse: though hasteners are usually causes, this isn t always the case The considerate doctor. A physician induces labour and thereby prevents the child from having to fight in the Vietnam war (because only men born between July 1 and December 31 are drafted) Lesson: we sometimes call hasteners causes, sometimes preventers, sometimes mere modifiers

32 Delayers The same is true of delayers:

33 Delayers The same is true of delayers: The evil doctor. The luckless draftee will blame (causally and/or morally) his having to serve in the war on the doctor s action or any other delayer if that s what caused the late birth

34 Delayers The same is true of delayers: The forest fire (Bennet 1987). The April rains did not cause the forest fire in June

35 Delayers The same is true of delayers: Saving lives. Most actions that delay someone s death (e.g. a physician s intervention) prevent that person s death

36 Normativity in Modifying It is clear that normative considerations play a role in determining what s a cause, for instance whether certain events are desirable or not and whether or not someone s action was appropriate Hasteners and delayers are more likely to be called causes when they involve intentional and morally relevant action Dying early is undesirable and therefore hastening is causing and delaying preventing; being born early is usually evaluatively neutral; but if being born late is undesirable, then a delayer can be a cause These cases show that pragmatic/contextual elements are required for more than just to determine contrast events: they also influence whether a factor is regarded as a cause, a preventer or a modifier

37 Final formulation and Conclusions Contextual Causation*. In X, C causally affects E if C, E are actual events; C*, E* are alternative, objectively possible events in X; and E* counterfactually depends on C*. X also determines the correct causal concept: causes/ prevents/affects the manner of (You may restore ambiguity by underspecifying the context) This concept of cause is relevant for history and the law Future work: making precise what is meant by X, objectively possible such as to be able to use the account for causal inference