A Revealed Preference Framework for AI Alignment

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autore principale: Suleymanov, Elchin
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914431368167424
author Suleymanov, Elchin
author_facet Suleymanov, Elchin
contents Human decision makers increasingly delegate choices to AI agents, raising a natural question: does the AI implement the human principal's preferences or pursue its own? To study this question using revealed preference techniques, I introduce the Luce Alignment Model, where the AI's choices are a mixture of two Luce rules, one reflecting the human's preferences and the other the AI's. I show that the AI's alignment (similarity of human and AI preferences) can be generically identified in two settings: the laboratory setting, where both human and AI choices are observed, and the field setting, where only AI choices are observed.
format Preprint
id arxiv_https___arxiv_org_abs_2603_27868
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle A Revealed Preference Framework for AI Alignment
Suleymanov, Elchin
Theoretical Economics
Artificial Intelligence
Computer Science and Game Theory
Human decision makers increasingly delegate choices to AI agents, raising a natural question: does the AI implement the human principal's preferences or pursue its own? To study this question using revealed preference techniques, I introduce the Luce Alignment Model, where the AI's choices are a mixture of two Luce rules, one reflecting the human's preferences and the other the AI's. I show that the AI's alignment (similarity of human and AI preferences) can be generically identified in two settings: the laboratory setting, where both human and AI choices are observed, and the field setting, where only AI choices are observed.
title A Revealed Preference Framework for AI Alignment
topic Theoretical Economics
Artificial Intelligence
Computer Science and Game Theory
url https://arxiv.org/abs/2603.27868