Integrating Causal Bayes Nets and inferentialism in conditional inference

This paper argues that recent developments in inferentialism in the psychology of reasoning that challenge the suppositional approach advocated by David Over can be implemented in Causal Bayes Nets (CBNs). Inferentialism proposes that conditionals, if p then q, imply (either as a matter of their meaning or a conventional implicature) that there is an inferential dependency between p and q. These dependencies can be captured in the directional links of a CBN (p → q), which can, therefore, provide a theory of mental representation and inference that inferentialism currently lacks. This approach has already been demonstrated for causal conditionals. We conclude that this proposal, while losing some inferences valid in the suppositional view, gains others that we know people make while also retaining consistency with the general Bayesian framework for human reasoning.