The alignment problem as a problem in metaethics.
A paper for the eleventh international Philosophy Students’ Symposium, “AI and Philosophy”, at the Faculty of Arts, University of Maribor, 12–13 November 2026. Keynotes by Atoosa Kasirzadeh (Carnegie Mellon and Google DeepMind) and Mario Günther (Panthéon-Sorbonne).
The paper argues that the dominant technical framing of AI value alignment treats human values as a latent signal to be recovered from data, that this quietly presupposes a contested metaethical picture, and that once the presupposition is made explicit the problem changes shape: it becomes a question of legitimacy rather than of correctness, and so belongs to political philosophy.
The dossier and an audio briefing are available on request — william@connelly-research.com.