Apt Causal Models and the Relativity of Actual Causation

Causal models provide a promising framework for analyzing actual causation. Such analyses must include how a model should map onto the world. While universally endorsed that a model must be accurate – saying only true things – the implications of this aren’t explored. I argue that, surprisingly, accuracy is not had by a model tout court, but only relative to a space of possibilities. This discovery raises a problem for extant causal model theories and, indeed, for any theory of actual causation in terms of counterfactual or type-level causal dependence. I conclude with a view that resolves this problem.