Ethical Nudging with Opaque Artificial Intelligence Systems?

Abstract

This paper examines the challenge to transparency that deep learning models used in recommender systems raise for ethical nudging. In particular, I will show how a lack of transparency may make it difficult to assess whether the nudges 1) make people better-off as judged by themselves and 2) do so without unduly interfering with their autonomy.

Philippe Verreault-Julien

Postdoctoral Researcher

Philosopher working on the ethics, epistemology, governance, and safety of artificial intelligence systems.