Abstract

Philosophers often appeal to Bayesian confirmation theory to render speculative metaphysical arguments more rigorous. But can a probabilistic framework designed for empirical data decide questions of metaphysics? This paper argues that Huemer’s Bayesian “existence → immortality” inference fails on two independent grounds. First, the comparison tacitly shifts the content of the evidence across the competing identity theories: treating “I exist now” as a different event under a restrictive view than under a permissive view. So, the Bayesian evidence is ill-defined. Second, if we hold the evidence fixed as the single indexical proposition E = “I exist now”, the key question is the relative likelihoods of E under each hypothesis. Once standard constraints on background verses evidence and on self-locating information are respected, there is no non-question-begging route to a Bayes factor that favors the permissive view of personal identity. Thus, mere existence does not confirm reincarnation. I remain agnostic on the correct theory of personal identity.