The other day George Bealer did his “Mental Causation” paper at MIT. It was fun, or at least as much fun as a 90 minute paper at the end of a long day can be. The main idea is rather clever – intentional actions follow by causal laws from both mental states (e.g. decisions) and physical states (e.g. neuron patterns) but the mental states get to count as causes because they “trump”, in Jonathan Schaffer’s sense, the physical states.

There’s plenty of examples that we could use to illustrate trumping. Here’s one that Bealer leans on a bit. The troops are trained to respond to any order from a higher ranking member of the army, and if conflicting orders come in, to respond to the order from the higher ranking of the orderers. The sargeant and the major both shout “Advance!” and the troops advance. What caused the troops to advance? Some people (not all!) have the intuition that it is really the major’s shout, not the sargeant’s. How could we capture that intuition? One nice suggestion is that it is because when we vary what the sargeant shouts, nothing changes in what the troops do, but if we vary what the major shouts, in general it does vary what the troops do. (‘In general’ rather than always, because if the major does not shout, there is no change in what the troops do.)

Induction on 1 case is a dangerous game, but that’s never stopped me from using it and it doesn’t stop Bealer either. Based on that case he formulates the following principle:

Assume c and d are determinables of action types, in particular actions by the entities that are prima facie plausible causes of the effects. Let c(x) and d(y) be the actions that are actually performed, b the actual background conditions, and e the actual effect. Then the following two conditions are sufficient for c(x) (rather than d(y)) being the cause of e.

(i) For most (typical) d(y’), in every nearest world in which original background b and c(x) occur, and d(y’) occurs instead of d(y), e still occurs.

(ii) For most (typical) c(x’), it is not the case that in each nearest world in which original background b and d(y) occur, and c(x’) occurs instead of c(x), e occurs.

I love counterfactual hypotheses about causation, because it’s such pleasant amusement coming up with counterexamples. The main trick is just to put backups everywhere, and bad things happen to the analysis. From then on it’s just a matter of adding the jokes, or at least making the example intrinsically amusing enough.

Assume that our army unit is mobbed up. They don’t do anything that The Mob Boss doesn’t approve of. It’s too dangerous for a Mob Boss to be involved in the day to day running of an army unit, so he delegates responsibility. For now the Mob Boss approves of the major, so the members of the army follow the major’s orders. The sargeant is known to be a favorite of the Mob Boss, so if the major ever loses authority, the soldiers know to follow the sargeant’s orders instead.

Today the Mob Boss is conducting his monthly review of the unit. At the end of the review, he always nods, indicating approval of the major, or shakes his head, indicating that the sargeant is to take over, and the major to be taken out and shot at the earliest convenient opportunity. Just as the review is ending, the following three things happen.

(a) The Mob Boss nods his head.
(b) The major shouts “Retreat!”
(c) The sargeant shouts “Retreat!”

It seems to me that the Mob Boss and the major are jointly causes of the ensuing retreat, and the sargeant’s order has little to do with it. But it does not seem right to say that the Major is the cause of the retreat – the troops are paying just as much attention to what the Mob Boss does as to the major’s shout, and only follow his order because of the nod.

Nonetheless, Bealer’s principle suggests that the major’s shout is the cause of the retreat. Since the sargeant is clearly out of the causal chain we can ignore his contribution. Now apply the above test, with c(x) being the major’s shout, and d(y) being the Mob Boss’s nod. If we vary d(y), the troops still retreat – they will shoot the major later, but only when it is safe to do so, and that’s not in the middle of a retreat. If we vary c(x), on the other hand, something other than a retreat happens. So according to Bealer, the major is the cause of the retreat.

I may have picked up this example from someone else. I remember talking to Simon Keller about counterexamples to principles like this one, and I’ve probably talked to other people without remembering that I was ‘borrowing’ ideas unacknowledged. Apologies if it’s your example and I haven’t said so. If it is my example, however, feel free to cite it wherever you like. I’m very unlikely to write this up, so this blog post will probably be the only citation source for it.