Newest papers

Reading Too Much Into It? Interpretive Preferences and Commitment in LLM Mentalisation under Ambiguity

Stephanie Fong, Yiwen Jiang, Qingyang Xu, Zimu Wang, Yaling Shen, Xiangyu Zhao, Jiahe Liu, Yinghao Zhang, Vincent Lee, Dominic Edmund Dwyer · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28105 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

Grandes modelos de linguagem (LLMs) são cada vez mais usados para aconselhamento interpessoal, potencialmente moldando como os usuários interpretam encontros sociais ambíguos. As avaliações existentes da Teoria da Mente geralmente dependem de respostas pré-definidas, enquanto a pesquisa de incerteza ignora as explicações preferidas e a confiança dos LLMs sob angústia psicológica. Para resolver essas lacunas, apresentamos o MentalLoom, uma nova avaliação da mentalidade do LLM que inclui 604 episódios sociais ambíguos, revelações de depressão e ansiedade social correspondentes e diálogos de quatro voltas com evidências de eventos inalterados. Ao longo de oito LLMs, as revelações iniciaram mais explicações negativas no geral, mas mais interpretações positivas direcionadas ao narrador.

Machine translation (Meta M2M100); the English original follows.

Abstract

Large Language Models (LLMs) are increasingly used for interpersonal advice, potentially shaping how users interpret ambiguous social encounters. Existing Theory of Mind evaluations typically rely on predefined answers, while uncertainty research overlooks LLMs' preferred explanations and confidence under psychological distress. To address these gaps, we introduce MentalLoom, a novel evaluation of LLM mentalisation comprising 604 ambiguous social episodes, matched depression and social anxiety disclosures, and four-turn dialogues with unchanged event evidence. Across eight LLMs, disclosures initially elicited more negative explanations overall but more positive narrator-directed interpretations. Scripted repeated questioning showed greater emotional support, stronger positive narrator-directed endorsement, and higher certainty without new event evidence. These findings show that user context can reshape interpretive preferences and commitment while the underlying ambiguity remains unresolved.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Ambiguity, Psychology, Certainty, Social psychology, Context (archaeology), Reading (process)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange