Framing the Machine: The Effect of Uncertainty Expressions and Presentation of Self on Trust in AI

Jacobsen, Cecilie Ellegaard ; Argot, Tania Camilla Taarsted ; Hansen, Emma Holtegaard

4. Term

Digitalisation and Application Development, MSc

2025

2025-06-04

14 pages

As Large Language Models (LLMs) become increasingly integrated into everyday life and work, understanding how their communication style influences user trust is critical. This study investigates how Uncertainty Expressions (Uncertain/Certain) and Presentation of Self (“I”/“The system”) affect user trust in LLM responses. Through a within-subjects 2×2 factorial design, 24 participants interacted with a chatbot across four experimental conditions while answering trivia questions. Trust, both perceived and demonstrated, was measured through validated questionnaires and behavioural indicators, with interviews conducted post-experiment to gather qualitative insights. Findings suggest that expressed certainty in chatbot responses significantly increases perceived Competence and the likelihood of selecting the chatbot as the primary source of information. Presentation of Self had a more nuanced effect, with first-person phrasing enhancing perceived Integrity for some users when uncertainty is expressed, while others preferred neutral, system-like framings. Additionally, participants often relied on Google’s top search result as a verification tool, highlighting the complex calibration of trust in human-AI interaction. Our results underscore the importance of designing LLM communication strategies that account for both linguistic cues and user context to foster and appropriately calibrate user trust in AI systems.

Download
View record in AAU Student Projects