Doomsday or Multiverse Bias

ABSTRACT Bayesian epistemology faces serious challenges when dealing with self‐locating evidence. This paper argues that, given two modest assumptions about how confirmation should work (Patterning and Live Centers), Bayesianism faces an unavoidable dilemma. We must either accept an implausible form of reasoning where our mere existence provides evidence that the future is short (Doomsday), or accept an equally implausible form where our existence favors hypotheses positing vastly more observers (Multiverse Bias). Since the underlying assumptions are hard to deny, and both horns of the dilemma are deeply unpalatable, this result reveals a major problem for Bayesian epistemology.