Year: 2024
Type: conference-paper
Abstract: It has become widely recognized that machine learning (ML) systems are value-laden. This raises a value demarcation problem: how can we distinguish between legitimate and illegitimate non-epistemic value influences in ML development and use? This paper makes two contributions. First, it surveys value demarcation strategies in ML and identifies gaps in the debate. Second, it addresses a deeper issue: what makes for a good demarcation strategy? We need a way to judge the adequacy of existing demarcation strategies across contexts. I submit contextual adequacy as a meta-norm for evaluating the prima facie justification of value demarcation proposals in ML.
Author: Streppel, Yeji
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Keywords: machine learning, metanorms, algorithmic ethics
Open Access status: green
OpenAlex ID: W7008988262
