arXiv cs.AI· Daniel Halpern, Evi Micha, Ariel D. Procaccia, Benjamin Schiffer, Itai Shapira, Shirley Zhang·· 2 天前AI 评分26
从社会选择理论视角看 AI 对齐
AI Alignment From Social Choice Perspectives
AI 导读
一项综述研究用社会选择理论审视基于人类反馈的对齐，指出当人类判断反映对理想行为的冲突观点时，学习到的目标就变成对模型偏好的聚合裁定。该研究梳理了近期工作，说明社会选择视角如何帮助识别反馈聚合层的失效模式，并揭示以显式、有原则的方式处理分歧的更广设计空间。
正文 · 原文
Abstract:Alignment from human feedback uses human judgments about model outputs to steer the behavior of language models after pretraining. When those judgments reflect conflicting views of desirable behavior, the learned objective becomes an aggregate determination of what the model should prefer. We survey recent work that has studied this aggregation problem through the lens of social choice theory. We illustrate how the social choice perspective helps identify failure modes in the feedback aggregation layer and reveals a broader design space for handling disagreement in explicit and principled ways.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2606.21550 [cs.AI] |
| (or arXiv:2606.21550v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2606.21550 arXiv-issued DOI via DataCite |
|
| Journal reference: | ACM SIGecom Exchanges 24(1):47-76, 2026 |
Submission history
From: Itai Shapira [view email]
[v1]
Fri, 19 Jun 2026 15:47:01 UTC (99 KB)
[v2]
Thu, 15 Oct 2026 15:34:05 UTC (51 KB)
来源：arXiv cs.AI · arxiv.org