Complex antecedents and probabilities in causal counterfactuals

Abstract

Ciardelli, Zhang, & Champollion [5] point out an empirical problem for theories of counterfactuals based on maximal similarity or minimal revision involving negated conjunctions in the antecedent. They also show that disjunctions and negated conjunctions behave differently in counterfactual antecedents, and propose an attractive solution that combines Inquisitive Semantics [4] with a theory of counterfactuals based on interventions on causal models [20]. This paper describes several incorrect empirical predictions of the resulting account, which point to a very general issue for interventionist theories: frequently the antecedent does not give us enough information to choose a unique intervention. The problem applies also to indefinites and to the negation of any non-binary variable. I argue that, when there are multiple ways of instantiating a counterfactual antecedent, we prefer scenarios that are more likely given general probabilistic causal knowledge. A theory is proposed which implements this idea while preserving [5]’s key contributions.