Do You Trust Me? Cognitive-Affective Signatures of Trustworthiness in Large Language Models

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Yeo, Gerard, Churina, Svetlana, Jaidka, Kokil
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866917206091104256
author Yeo, Gerard
Churina, Svetlana
Jaidka, Kokil
author_facet Yeo, Gerard
Churina, Svetlana
Jaidka, Kokil
contents Perceived trustworthiness underpins how users navigate online information, yet it remains unclear whether large language models (LLMs),increasingly embedded in search, recommendation, and conversational systems, represent this construct in psychologically coherent ways. We analyze how instruction-tuned LLMs (Llama 3.1 8B, Qwen 2.5 7B, Mistral 7B) encode perceived trustworthiness in web-like narratives using the PEACE-Reviews dataset annotated for cognitive appraisals, emotions, and behavioral intentions. Across models, systematic layer- and head-level activation differences distinguish high- from low-trust texts, revealing that trust cues are implicitly encoded during pretraining. Probing analyses show linearly de-codable trust signals and fine-tuning effects that refine rather than restructure these representations. Strongest associations emerge with appraisals of fairness, certainty, and accountability-self -- dimensions central to human trust formation online. These findings demonstrate that modern LLMs internalize psychologically grounded trust signals without explicit supervision, offering a representational foundation for designing credible, transparent, and trust-worthy AI systems in the web ecosystem. Code and appendix are available at: https://github.com/GerardYeo/TrustworthinessLLM.
format Preprint
id arxiv_https___arxiv_org_abs_2601_10719
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Do You Trust Me? Cognitive-Affective Signatures of Trustworthiness in Large Language Models
Yeo, Gerard
Churina, Svetlana
Jaidka, Kokil
Artificial Intelligence
Computation and Language
Perceived trustworthiness underpins how users navigate online information, yet it remains unclear whether large language models (LLMs),increasingly embedded in search, recommendation, and conversational systems, represent this construct in psychologically coherent ways. We analyze how instruction-tuned LLMs (Llama 3.1 8B, Qwen 2.5 7B, Mistral 7B) encode perceived trustworthiness in web-like narratives using the PEACE-Reviews dataset annotated for cognitive appraisals, emotions, and behavioral intentions. Across models, systematic layer- and head-level activation differences distinguish high- from low-trust texts, revealing that trust cues are implicitly encoded during pretraining. Probing analyses show linearly de-codable trust signals and fine-tuning effects that refine rather than restructure these representations. Strongest associations emerge with appraisals of fairness, certainty, and accountability-self -- dimensions central to human trust formation online. These findings demonstrate that modern LLMs internalize psychologically grounded trust signals without explicit supervision, offering a representational foundation for designing credible, transparent, and trust-worthy AI systems in the web ecosystem. Code and appendix are available at: https://github.com/GerardYeo/TrustworthinessLLM.
title Do You Trust Me? Cognitive-Affective Signatures of Trustworthiness in Large Language Models
topic Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2601.10719