On expected utility, part 3: VNM, separability, and more

Abstract

This article considers the philosophical argument that we should be assigning significant credence to the hypothesis that we are living in a computer simulation. It distinguishes between two types of simulation arguments: Type 1 arguments, which rely on our standard story about the world to show that it is unlikely that we are not living in a simulation, and Type 2 arguments, which rely instead on a specific claim about reasonable priors over structurally similar worlds. The article argues that Type 1 arguments are flawed, both because they are subject to counterexamples and because they rely on an overly narrow conception of what counts as “knowing” something. In contrast to Type 1 arguments, Type 2 arguments are less vulnerable to these objections and provide a more solid basis for assigning credence to the hypothesis that we are living in a simulation. However, they also face a number of challenges, including the difficulty of assessing “structural similarity” and the risk of committing an “epistemic pascal’s mugging.” Overall, the article argues that while simulation arguments are not decisive, they do provide some reason for taking seriously the possibility that we are living in a simulation and that further research on this topic is warranted. – AI-generated abstract.