| Summary: | <p>Building on the work of David Albert in the foundations of statistical mechanics, Barry Loewer has developed a novel analysis of counterfactuals within a powerful system for naturalized metaphysics that he and Albert call "the Mentaculus". In this thesis, I apply the Mentaculus to the problem of causation. In order to relate Loewer's work on counterfactuals to an account of causation, I take my cue from the framework of structural causal models (SCMs) and the associated methods of causal-statistical inference. I explain how this framework is not immediately congenial to the Mentaculus, since the former is poorly placed to handle counterfactuals with probabilistic consequents, which are central in the latter. To solve this problem, I offer a natural generalization of the formalism of SCMs to the indeterministic setting. I then use the generalized formalism to formulate a criterion for causal influence in terms of patterns of counterfactual dependence (the "intervention criterion"). Putting this criterion together with Loewer's analysis of counterfactuals, I describe a method for reconstructing a causal graph using the Mentaculus.</p>
<p>The payoff for this effort is twofold. First, by embedding the (time-neutral) framework of causal models within the time-asymmetric system of the Mentaculus, an explanation emerges as to why causal models can be expected to align themselves with the direction of time, with causes always preceding their effects in time. Second, by showing how a causal graph may be <em>reconstructed</em> out of the Mentaculus, a number of common assumptions about causality---usually presupposed in the causal-models formalism---can be vindicated. These include not only the presupposition of a directed and acyclic structure for the causal graph, but also the truth of a contentious principle relating causal structure to patterns of statistical independence.</p>
|
|---|