Preemption in Singular Causation Judgments - A Computational Model (Topics version)

Abstract

Causal queries about singular cases are ubiquitous, yet the question of how we assess whether a particular outcome was actually caused by a specific potential cause turns out to be difficult to answer. Relying on the causal power framework (Cheng, 1997), Cheng and Novick (2005) proposed a model of causal attribution intended to help answer this question. We challenge this model, both conceptually and empirically. We argue that the central problem of this model is that it treats causal powers that are probabilistically sufficient to generate the effect on a particular occasion as actual causes of the effect, and thus neglects that sufficient causal powers can be preempted in their efficacy. Also, the model does not take into account that reasoners incorporate uncertainty about the underlying general causal structure and strength of causes when making causal inferences. We propose a new measure of causal attribution and embed it into the structure induction model of singular causation (SISC; Stephan & Waldmann, 2016). Two experiments support the model.

Publication
Topics in Cognitive Science, 10
“Best of Papers from the Cognitive Science Society Annual Conference,” Wayne D. Gray (Topic Editor). For a full listing of topic papers, see http://onlinelibrary.wiley.com/journal/10.1111/(ISSN)1756-8765/earlyview
Avatar
Simon Stephan
Research Scientist in the field of Cognitive Science at the

My research interest is computational cognitive science. I’m particularly active in the field of causal learning and causal reasoning.