arXiv cs.CL· Michele Loi·· 10 天前AI 评分22
为 AI 制定"认知宪法"：如何避免大语言模型的连贯性偏差
Epistemic Constitutionalism Or: how to avoid coherence bias
AI 导读
论文主张为 AI 建立"认知宪法"，即明确且可争议的元规范，用以约束大语言模型如何形成和表达信念。一项预注册研究发现来源标注对论证评估存在内容依赖效应，而来源独立性并非中立默认——在证言语境中，来源立场与逆利益发声的成本都可提供相关证据。作者区分了柏拉图式与自由主义两种设计路径，支持后者，并提出包含八项原则与四个取向的宪法核心。
正文 · 原文
Abstract:Large language models increasingly function as artificial reasoners: they evaluate arguments, assign credibility, and express confidence. Yet their responses can leave the epistemic policies governing these evaluations implicit. This paper argues for an epistemic constitution for AI: explicit, contestable meta-norms regulating how systems form and express beliefs. Source attribution provides the motivating case. An exploratory audit suggested that expectations about a source's position intrude on argument evaluation. A preregistered study (arXiv:2609.35286) then found content-dependent effects of source attribution, with selected written evaluations supporting source-position fit as an explanation. The audit also revealed conflicting justifications for attending to sources. Source independence, however, is not a neutral default: in testimonial contexts, a source's position and the costs of speaking against interest can provide relevant evidence. I distinguish two approaches to epistemic constitution design: the Platonic, which mandates formal correctness and default source-independence from a privileged standpoint, and the Liberal, which rejects such privilege and protects conditions for collective inquiry while allowing principled source-attending grounded in epistemic vigilance. I defend the Liberal approach, sketch a constitutional core of eight principles and four orientations, and argue that AI epistemic governance requires explicit, contestable norms for evaluating testimony, responding to evidence, and revising judgements.
| Comments: | 33 pages, 1 table. Substantial revision: empirical discussion updated in light of arXiv:2609.35286; exploratory-audit claims corrected against public logs. Appendix A: full per-log register. Appendix B: corrections to v4 and AI-assisted writing documentation. Philosophical argument clarified |
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computers and Society (cs.CY) |
| Cite as: | arXiv:2601.14295 [cs.AI] |
| (or arXiv:2601.14295v5 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2601.14295 arXiv-issued DOI via DataCite |
Submission history
From: Michele Loi Dr. [view email]
[v1]
Fri, 16 Jan 2026 07:36:30 UTC (4,210 KB)
[v2]
Tue, 27 Jan 2026 19:15:46 UTC (4,216 KB)
[v3]
Wed, 22 Apr 2026 11:33:20 UTC (549 KB)
[v4]
Thu, 11 Jun 2026 09:22:30 UTC (773 KB)
[v5]
Tue, 6 Oct 2026 23:13:18 UTC (38 KB)
来源：arXiv cs.CL · arxiv.org