Self-locating credences

Thesis

In this essay, I will argue for a theory of self-locating credences based on Nick Bostrom’s Observation Equation. A theory of self-locating credences should constrain a rational observer’s credence function given their prior credences in non-self-locating propositions and their beliefs about how many observers exist in the world and what kind of evidence they have. Two notable theories of this kind have been proposed: one by Frank Arntzenius and Cian Dorr, and another by Nick Bostrom. In the essay, I argue that Arntzenius and Dorr’s theory of self-locating credences has strongly counter-intuitive results when applied to certain kinds of cases, as Nick Bostrom has shown. On the other hand, Bostrom’s theory is incomplete—Bostrom says that observers should be separated into reference classes in order to avoid the results that a theory like Arntzenius and Dorr’s gives, but does not say anything about how observers should be separated into reference classes. In this essay, I develop a principled reference class definition to complement Bostrom’s theory. I first give a preliminary reference class definition that is adequate in a special case—namely, whenever every observer knows what reference class they are in according to this preliminary definition. I then formulate a general reference class definition that I claim works generally, even when observers do not know what reference class they are in according to the preliminary definition. Following this, I apply the reference class definition to the Sleeping Beauty problem and conclude that her credence in heads should be 1/2 throughout. In the last part of the essay, I defend the theory from some potential objections.

Authors: Ivankovic, J

• DOI: 10.5287/ora-jbgaje67p
• Type of award: BPhil
• Level of award: Bachelors
• Awarding institution: University of Oxford
• Language: English