Machine Learning and Theory Ladenness -- A Phenomenological Account

TL;DR

This paper analyzes how machine learning models relate to scientific domain theories, arguing that ML is largely indifferent to domain knowledge, which impacts their transferability and challenges existing views on theory ladenness.

Contribution

It offers a phenomenological account showing ML models are mostly indifferent to domain theories, contrasting with recent philosophical trends and highlighting implications for scientific transferability.

Findings

ML models are mostly indifferent to domain theory

ML model-building exhibits weak theory ladenness, termed 'theory infection'

Implications for transferability of ML across disciplines

Abstract

We provide an analysis of theory ladenness in machine learning in science, where "theory", that we call "domain theory", refers to the domain knowledge of the scientific discipline where ML is used. By constructing an account of ML models based on a comparison with phenomenological models, we show, against recent trends in philosophy of science, that ML model-building is mostly indifferent to domain theory, even if the model remains theory laden in a weak sense, which we call theory infection. These claims, we argue, have far-reaching consequences for the transferability of ML across scientific disciplines, and shift the priorities of the debate on theory ladenness in ML from descriptive to normative.