Description
In our next meeting, we will discuss the article "Moral disagreement and the limits of AI value alignment: a dual challenge of epistemic justification and political legitimacy" by Nick Schuster & Daniel Kilov. The article argues that current approaches to AI value alignment—such as crowdsourcing, reinforcement learning from human feedback, and constitutional AI—do not adequately address reasonable moral disagreement when AI systems make morally controversial decisions. It concludes that because these methods fail to provide sufficient epistemic and political justification for affected stakeholders, accommodating moral pluralism remains an unresolved challenge for AI safety research.
Associate Professor Michael Barnes at the Department of Philosophy, Classics, History of Art and Ideas (IFIKK) at UiO will open the session with a short introduction to the article, outlining its main arguments.
Participants are strongly encouraged to read the article in advance and to come prepared with questions, comments, and critical reflections, as the session is intended to be discussion-driven.
Resources
Schuster, N., Kilov, D. Moral disagreement and the limits of AI value alignment: a dual challenge of epistemic justification and political legitimacy. AI & Soc 40, 6073–6087 (2025). https://doi.org/10.1007/s00146-025-02427-2
Practical
This is an open meeting held both in person and online. For online attendance: https://uio.zoom.us/j/62960689968
This reading club is organised as part of research theme on Ethics (RT9)