24.221 Metaphysics Thursday November 17, 2005 Sufficient reason. Problem for the causal network approach. I call it the “faux hero” problem, after the firefighter who is there to put the fire out only because she started it.Suppose c the my placing the bomb under your chair; d is your fleeing upon spotting thebomb; b is the chair’s blowing up; e is your surviving. Here’s the graph, annotated toclarify the dependence relations: D B E = D v~BC The path from C to E through B is inactive, since holding off-path variable D fixed at itsactual value 1, E is 1 regardless of what value C takes. But the path through D is active,since holding B fixed at its actual value 1, E is 1 iff C = 1. So the account wrongly predictsthat my faux-heroic act of putting a bomb under your chair caused your survival, by tippingyou off to the impending explosion. How should the network approach deal with this? Unintuitive. “The interpolated variable B is not easy to find, and the …counterfactual thatreveals the active caudsal route from [C to E] is not at all intuitive. [For] the relevant piece ofcounterfactual reasoning would go as folows: Suppose that the [chair had blown up], andsuppose moreoever that that [there had been no bomb under it] in the first place. Since [therewas no bomb, you] would not have seen [the explosion] coming and would not have [runaway]; it would have [occurred right under you] and [you] would not have survived…. This is the sort of counterfactual reasoning that only trained philosophers engage in; unaidedintuition is not to be faulted for failing to “see” the relevant…counterfactual.” Reply. This seems less like a refutation of the imagined trained philosopher than an apologia on behalf ofthe poor non-philosopher. But the non-philosopher, if she denies c causes e, is correct! Far-fetched. The above figure “does not constitute an appropriate representation” of[BOMB]….our causal judgments depend on which unactualized possibilities we are willingto take seriously, and which we consider too remote. The variables we choose to include should reflect our concerns…When we exclude the variable B from our model…[that isbecause] we are not willing to take seriously the possibility that the [chair explodes eventhough there is no bomb under it]. This possibility is just too far fetched.” Reply. Changethe example a bit; C is my pushing a delicately balanced bomb off a ledge to fall on yourchair; D is your noticing my doing this and fleeing; B is the chair’s exploding; E is yoursurvival. It’s not at all plausible that my action was a cause of your survival. But it’s not at all far-fetched that the delicately balanced bomb would fall for some other reason.Sufficient reason (1). Distinguish a variable V’s taking a default value from its taking adeviant value. Intuitively, default values are what you’d expect anyway; deviant values depart from the norm in a way that intuitively requires explanation. The principle ofsufficient reason says that a variable V with parents V1…Vn does not take a deviant value unless one of the Vis takes a deviant value. Call a network self-contained if every variable inthe network satisfies the principle of sufficient reason. “The main idea I will defend [calledTC for token causation] is that counterfactual dependence [dependence not holding anythingfixed] is necessary and sufficient for token causation in self-contained networks.” Example. C = my placing the bomb. D = your fleeing. E is your surviving. D = C and E = D v ~C. This network is self-contained (why?). So we look for straight counterfactualdependence, not holding anything fixed. There is none: E = 1 if C = 1 because then D = 1,and E = 1 if C = 0 because ~C is a disjunct of E. So we now have a story about why the fauxhero is not a cause. Complaint. The network account’s pride and joy is standard early preemption cases, e.g.,Suzy throwing so that Billy doesn’t bother. But the causal network here is not self-contained, for B takes a deviant value (1 for Billy-throw) when its lone parent S assumes itsdefault value (0 for Suzy-no-throw). This means some of our strongest positive judgmentsare no longer confirmed by the theory. Sufficient reason (2). “The reader cannot help but have noticed that TC provides a necessarycondition for causation, and a sufficient condition, but leaves a gap between them. What are we to say about cases where the causal network is not self-contained? …What I suspect isthat something like the [active-route account] is essentially correct, but that the verdicts ofsuch an account can be partially or completely counteracted by TC when the latter yields aclear negative conclusion.” The suggestion then seems to be this: c causes e iff EITHER the network is self-contained and e depends on c absolutely, that is holding nothing fixed (notesufficient reason is automatically satisfied), or it is not self-contained and e depends on c holding off-path variables fixed at their default values (so as to minimize violations ofsufficient reason). Reply. Suppose Bodyguard inserts antidote (D =1) iff he sees Assassin inserting poison (A=1). Then Victim survives (E = 1) no matter what. Intuitively what Bodyguard does savesVictim’s life; it’s a cause of Victim’s survival. But the network is self-contained (why?).And we saw there is no counterfactual dependence; Victim survives no matter what.Conversely we can change BOMB so the network is not self-contained; the bomb is droppedbecause I remove the note telling my assistant not to drop it. Still seems like faux heroism,but now we’re told causation goes with an active path, and there is one; holding fixed that thechair was going to explode, you survive iff you are tipped off by my removing the