Non-Rational Belief Paradoxes as Byzantine Failures

Auto-epistemology (updating principles like Reflection and Re-Cal) and collective epistemology (group belief, peer disagreement, etc.) are both hot topics in formal epistemology. Christensen (1991) linked these two areas by showing that synchronic Dutch books against groups with differing credences had the same properties as diachronic Dutch strategies against individuals with differing credences across time. Christensen’s conclusion is that both are suspect—neither groups at a time nor individuals across time should be held to strict epistemic consistency, and thus Dutch strategies should be rejected as a binding constraint on rationality. This conclusion is problematic, however, because Dutch strategies are the critical arguments linking Conditionalization to the Kolmogorov axioms. I dispute Christensen’s rejection of Dutch strategies by characterizing them as instances of what computer scientists call Byzantine failures (Lamport et al., 1982). This characterization supports Christensen’s comparison between synchronic Dutch books against groups and diachronic Dutch strategies against individuals. The same underlying faults that cause Byzantine failures can also cause susceptibility to Dutch books, however, so Dutch strategies do not merit suspicion. Rather, agents (and groups) susceptible to Byzantine failures cannot count as ideally rational.

Bio:

Ryan is a National Science Foundation Doc.CH fellow in the Metaphysics of Quantum Objects group at the University of Geneva, working on grounding and parthood in realist interpretations of relativistic quantum mechanics. He has taught both at Johns Hopkins University and in Papua New Guinea. Having worked at Google and TripAdvisor between his undergraduate and graduate studies, Ryan is still fascinated by the philosophical issues surrounding computation, including Turing machines, artificial intelligence, and conceptual engineering in software development. He also really loves prata.