Abstract
In Making Things Happen, I claimed that it is methodologically useful to interpret causal claims as claims about what would happen under interventions. This "interventionist" idea has been criticized on a number of grounds. Among other considerations, critics have maintained that (1a) interventionism is unhelpful in cases involving causal inference from non-experimental data, that (2a) contrary to what I have claimed, interventionism is irrelevant to methodology, that (3a) interventionism is "circular" (because the notion of an intervention is itself a causal notion) and that this makes it unilluminating or illegitimate as a treatment of causation, that (4a) interventionism is defective because it fails to provide an acceptable account of the "truth conditions" for causal claims, and that (5a) its reliance on the notion of "possible interventions" is both unclear and unnecessary.
In this talk, I will attempt to respond to these criticisms and to clarify what I take to be the key elements of interventionism. I will suggest, first, that (1b) an interventionist conception of causation can play an important role in characterizing the target to which we are trying to infer when WE make causal inferences in non-experimental contexts: Roughly, in non-experimental contexts we are trying to infer what the results of a hypothetical experiment would be without doing the experiment). We can use this "target" idea to explain the utility and reliability of certain well-known inferential techniques such as instrumental variables and regression discontinuity designs, thus demonstrating one important respect in which interventionism is methodologically fruitful. I will also argue that (2b) interventionist ideas are methodologically useful for other reasons as well: for example, because they have implications for which variables can figure in causal claims. These considerations will lead to (3b) an exploration of the general question of how it can be illuminating, as the interventionist supposes, to appeal to (interventionist) counterfactuals that themselves presuppose causal ideas, in order to elucidate the content of causal claims and the methodology of testing them will also be explored. I believe that the answer to this question has both a logical/methodological and perhaps a psychological aspect. As an empirical matter, practitioners in a number of disciplines seem to find accounts of this sort illuminating despite their "circularity"- the questions for philosophers should be how this is possible and what the illumination in question consists in. Finally with respect to (5a), I will argue that (5b) although it may seem initially puzzling to include reference to the possibility of intervening on C in elucidating causal claims like C causes E, in fact there are a number of real-life examples of causal reasoning in which we seem (justifiably) to do exactly this. Time permitting, I will also make some observations about the project of providing "truth conditions" for causal claims - what this might mean and for what purposes it might be desirable.
