2026-10-27 · ArXiv
arXiv:2601.14271v4 Announce Type: replace
Abstract: Shared accountability records must often remain useful to parties who disagree about causation, responsibility, and normative interpretation. Suppressing these claims would erase information needed for accountability; asserting one disputed interpretation as foundational would privilege it. A neutral substrate instead separates shared referential commitments from attributed interpretive assertions.
This paper formalizes that separation through admissible frameworks, referential commitments, permitted attributions, interpretive non-commitment, and extension stability. Under Referential Common Ground, restricting every substrate-layer commitment to a referential commitment or a permitted attribution proposition is sufficient for semantic neutrality relative to every admissible framework. It is not a design-time guarantee: such a guarantee requires a permitted basis establishing the neutrality result, rather than only its truth.
The essential assumption is joint compatibility of the complete referential commitment set with each admissible framework. Separate compatibility of individual commitments is not sufficient, and admissible frameworks need not be mutually consistent. Furthermore, the formal referential interfaces do not independently establish the identity criteria, denotation relations, attributional bases, or interpretive adequacy required by a particular application. When those premises cannot be justified, neutrality cannot be certified by the present construction.
The contribution is a precise conditional layer restriction and an explicit account of its semantic and scientific boundaries. Attribution and provenance are established representational mechanisms; their presence alone cannot establish referential fixing.
原文链接：https://arxiv.org/abs/2601.14271
查看更多资讯请见 AI 行业资讯。