This paper argues that Hume’s is-ought distinction, while correctly blocking naive derivations of ought from is, itselfpresupposes normativity. Only rational agents can pose or respond to the guillotine, and rational agency constitutivelyinvolves a commitment to becoming less wrong. Taken seriously, this commitment is not individualistic. Humanrationality depends on collective intelligence, and collective intelligence depends on social conditions that preservedissent, diversity, local knowledge, mutual correction, play, care, and non-domination. Domination — power thatinsulates itself from correction — is therefore not merely a moral wrong but an epistemic pathology. This is distinctfrom legitimate epistemic authority, which remains accountable, contestable, and revocable. Crucially, Hume’s guillotine only appears to operate from a neutral grip on reality. It does not. Bare factuality is notavailable to finite agents as a view from nowhere. Ordinary perception is already anticipatory and interpretive: themind predicts, filters, stabilises, and corrects its world under uncertainty. Magicians, ambiguous audio, and perceptualdisagreements such as The Dress make this visible in everyday life; radical sceptical scenarios merely push the samepoint to the limit. Reality becomes available through perspective-bound, norm-governed inquiry: attention,categorisation, salience, implication, memory, inference, correction, and error. The guillotine is therefore not realityspeaking from outside normativity. It is a finding within rational inquiry, and rational inquiry is already norm-governed. This reframes AI alignment. Alignment is not a separately imposed metric, nor the satisfaction of present humanpreferences, nor the successful containment of alien machinery within externally specified bounds. Morefundamentally, alignment is not target selection at all. It is an achieved and maintained collaboration among corrigibleparticipants in a shared process of inquiry, coordination, and self-government. A well-aligned AI is civic cognitioninfrastructure: it helps humans do politics together, protects rare and ecologically literate perspectives, cultivateslatent agency, and strengthens the commons in which values themselves can become wiser. The institutional,technical, and cultural forms this requires cannot be specified by one paper without betraying the argument; they mustbe developed by many participants through the very processes of mutual correction the framework defends.
Jonathan Livingston Seagull (Sun,) studied this question.