Pascal’s Mugging as a Challenge to Expected Utility Theory

Author’s Note

What was your thesis topic?

This thesis explores the claim that expected value theory is vulnerable to fanaticism – it allows options with extremely small probability of astronomically large value to overwhelm our decision, causing us to make counterintuitive choices. I reject the claim that expected utility theory is rendered redundant by fanaticism. I first examine the foundations of expected utility theory, investigating whether it must necessarily lead to the fanatic conclusion. I show that expected utiltiy theory cannot be discarded without entailing various issues. I propose the counterfactual solution: when conditioning on extremely small probabilities, we find that we do not have to be “mugged” by spurious decisions, because the opportunity cost is high. This cancels out the expected utility of fanatic outcomes. Thus, expected utility theory is not effectively challenged by Pascal’s mugging. I conclude by illustrating the relevance of this solution to the expected utility of reducing existential risk.

What do you think the stronger and weaker parts of your thesis are?

I believe my paper is strongest in laying out the foundations of expected utility theory, and highlighting the challenges that come with rejecting any of its premises. I think it makes an original contribution, by pointing out that these premises do not necessarily lead to the conclusion of a Pascalian mugging in the real world, and I argue that this is an acceptable outcome.

My conclusion suffers from the challenge of cluelessness. The reality is that many decisions have evidential asymmetry, meaning we face cases of complex (not simple) cluelessness. This means we can’t neatly “cancel out” fanatical options easily. Thus, in practice we still may be vulnerable to fanatical options in our decision set. I think this would be most important to explore with further research.

In what ways do you think your topic improves the world?

My paper shows that Pascal’s Mugging is not a valid critique to longtermism, since the arguments for expected utility theory are extremely strong. However, I also show that it is important to not make naive expected value calculations in practice when making decisions within a longtermist framework. This is important for those making practical cause prioritisation decisions such as donation and career choice.