arXiv cs.AI· Stine Lyngs{\o} Beltoft, Jonas Vistrup, Peter Schneider-Kamp, Lukas Galke Poech·· 2 小时前AI 评分31
主权基础模型的活宪法：参与式对齐研究论文提出五项制定原则
Living constitutions: Participatory alignment for sovereign foundation models
AI 导读
一篇 arXiv 论文提出为主权基础模型制定“活宪法”，通过类似公民科学的参与式过程进行迭代起草、测试与修订，并给出迭代性、模块化、可论证、立场性和正当性五项原则。条款可标记为已定、有争议或未定，将持久分歧记录在宪法中而非强行消除。基于智能体的建模显示，在审议群体中 75% 参与者善意的假设下，该流程可容忍 50% 的恶意参与者。
正文 · 原文
Abstract:In Constitutional AI, a plain-text "constitution" encodes values and gets incorporated into a model via post-training. However, the procedure for creating such a constitution poses significant challenges: who contributes, who decides which norms are included, and to what extent must the process be participatory or democratic? These challenges can be paralyzing. Yet, we argue that taking action is essential: without a deliberate constitution, sovereign models would risk misalignment or default to norms set by frontier model providers. We propose treating the development of a constitution as an evolving, participatory process akin to citizen science, iteratively drafted, tested, and refined. We therefore formulate five desirable principles: iterative, modular, justifiable, positional, and legitimate. We further argue that such a document must remain responsive not only over time but across the people it governs, and propose that provisions be marked as settled, contested, or underdetermined, so that durable disagreement is recorded in the constitution rather than resolved away. Empirically, we show via agent-based modelling that proposed process for maintaining living constitutions is robust against 50% bad actors among citizens - assuming a share of 75% participants with good intent in the deliberative body. This indicates that the proposed process tolerates a substantial share of adversarial participants.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.13402 [cs.AI] |
| (or arXiv:2611.13402v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.13402 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Stine Beltoft [view email]
[v1]
Tue, 10 Nov 2026 00:32:24 UTC (848 KB)
来源：arXiv cs.AI · arxiv.org