Home | Publications | Ste26a

Values in Machine Learning: What Follows From Underdetermination?

MCML Authors

Tom Sterkenburg

Dr.

Associated JRG Leader Epistemology in ML

Epistemological Foundations of Machine Learning

Abstract

It has been argued that inductive underdetermination entails that machine learning algorithms must be value-laden. This paper draws from the philosophy of induction to rather highlight the epistemic motivations and justifications that play a role in machine learning algorithm design. The analysis offered indicates that some of the arguments from underdetermination to value-ladenness are too quick, but it also supports their conclusion by indicating how the practical realization of these epistemic considerations inevitably introduces various non-epistemically value-laden judgments, too. The suggestion is that exposing value-ladenness is not inconsistent with, and even profits from, appreciation of the epistemic considerations involved.

article Ste26a