A Revealed Preference Framework for AI Alignment
Not on the map. Why?
Abstract (arXiv)
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.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 12 of its 27 references that are in the map. First posted on arXiv 29 Mar 2026.
Builds on
12 of 27 references.
References · 12
- 2017Deep reinforcement learning from human preferencescited by 3,912
- 2022Constitutional AI: Harmlessness from AI Feedbackcited by 3,050
- 2016Concrete Problems in AI Safetycited by 1,642
- 2022Discovering Language Model Behaviors with Model-Written Evaluationscited by 836
- 2018Scalable agent alignment via reward modeling: a research directioncited by 447
- 2024Alignment faking in large language modelscited by 360
- 2016Cooperative Inverse Reinforcement Learningcited by 300
- 2023AI Alignment: A Comprehensive Surveycited by 263
- 2023An Overview of Catastrophic AI Riskscited by 222
- 2022Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networkscited by 129
- 2022On the identifiability of mixtures of ranking modelscited by 7
- 2020Learning an arbitrary mixture of two multinomial logitscited by 3
Built on by
Citing papers per year: 2026: 1.
1 paper.