Monday, January 21, 2008

Consequentialism and counterfactuals

According to standard act consequentialist theories, an action is right if and only if that there is no alternate action within one's power do that would in fact have better consequences. Focus on the words "would in fact". Here we have a counterfactual. Moreover, it is a counterfactual where the consequent depends indeterministically on the antecedent. But suppose that one denies Molinism, and more generally denies that there can be any non-trivial counterfactuals where the consequent depends indeterministically (either via libertarian-free actions or through quantum randomness) on the antecedent. Then the act consequentialist theory cannot work.

One might say that our actions concern a subset of the world that we may assume is deterministic. But remember that standard consequentialist theories involve also the weighing of distant consequences. It is highly likely that between the present and a distant future, indeterministic events will have some quite significant consequences. It seems pretty likely that over a long enough period of time, for instance, there will be some car crashes for indeterministic causes (e.g., indeterministic effects in the brains of drivers, or quantum effects in defective engine-control electronics, or the like). Moreover, we surely shouldn't assume something false. If we accept quantum indeterminism, then strictly speaking all the stuff around is indeterministic, though it may have extremely high probability. But extremely high probability won't help those worried about whether there are non-trivial counterfactuals involving stochastic dependence.

Suppose one bites the bullet. One denies that there are any true counterfactuals about future results, but one accepts the analysis of rightness. Then one gets the result that every action is right. For no action is such that there is an action that would have better consequences, since there are not enough facts to make such a "would" true.

If we take this criticism seriously, we will either abandon consequentialism, or define rightness not in terms of what it is true to say "would happen", but in terms of expected values of actual and counterfactual outcomes. There is still a problem, though, whether it makes sense to talk of the expected values of counterfactual outcomes when one believes that there is no such thing as a "counterfactual outcome", as the typical Molinist does. One might be able to define the expected values in terms of present tendencies, but now the theory is sounding less and less like consequentialism.[note 1]


Heath White said...

The consequentialist doesn’t need to affirm that particular choices will lead to particular definite outcomes, i.e. counterfactuals of the form, “If I were to do A, Z would happen” for definite Z. What she needs is a relation, something like “for all actions in my power, there is some action A such that the outcome of A is better than the outcome of the other actions.” That is, the outcomes can be indeterministic or unknowable, but their relative values need to be deterministic or knowable. One might very well wonder, however, how the relative values of outcomes can be deterministic when the outcomes themselves are not deterministic.

Alexander R Pruss said...


I think your final sentence says it well. The reasons we have for doubting the determinacy of outcomes are reasons for doubting the determinacy of the values of outcomes. Moreover, some of them also seem to be reasons for doubting the determinacy of the relative values of outcomes.