Ambiguity Collapse by LLMs: A Taxonomy of Epistemic Risks

Researchers have identified 'ambiguity collapse' as a significant epistemic risk when large language models encounter ambiguous terms and produce singular interpretations without human deliberation. The phenomenon threatens decision-making processes in content moderation, hiring, and AI self-regulation by bypassing normal human practices of meaning negotiation and potentially distorting shared vocabularies over time.

• →LLMs increasingly make decisions on disputed concepts like 'hate speech' and 'qualified' candidates, creating new epistemic risks.
• →Ambiguity collapse occurs when AI systems resolve genuinely ambiguous terms into singular interpretations without human deliberation.
• →The phenomenon poses risks at three levels: process, output, and ecosystem-wide impacts on shared vocabularies.
• →Applications span content moderation, hiring decisions, and AI constitutional self-regulation systems.
• →Researchers propose multi-layer mitigation strategies including training modifications and interface design changes.