Published: Theory choice, non-epistemic values, and machine learning

In this academic paper, I use a theorem from machine learning to support the claim that we can’t choose between theories without relying on non-epistemic values, such as social and political values.


The paper was published by Synthese, a top philosophy journal. It also won the Fink Prize, for the best paper written by a UC Berkeley philosophy graduate student in 2019.


Here is the free preprint and here is the paywalled official version


This is a summary of the paper:


Can we choose between theories without relying on non-epistemic values, such as social and political values? I use a theorem from machine learning, called the “No Free Lunch” theorem (NFL), to support the claim that we can’t.


First, I argue that NFL entails that predictive accuracy is insufficient to favor a given theory over others. Second, I argue that NFL challenges our ability to give a purely epistemic justification for using other traditional epistemic virtues, such as simplicity or explainability, in theory choice. Third, I argue that the natural way to overcome NFL’s challenge is to use non-epistemic values in theory choice. Last, I argue that, contrary to common conception, the epistemic challenge arising from NFL is distinct from Hume’s problem of induction and other forms of underdetermination.

Recent Posts

See All

How can AI cause harm, and which harms are the most important? In particular, which AI risks are end-users most concerned about? Results fro