Blissful (A)Ignorance: People form overly positive impressions of others based on their written messages, despite wide-scale adoption of Generative AI

Authors: Jiaqi Zhua, Andras Molnar

Published: 2025

Publication: ArXiv

Impressions of written messages are overly positive when recipients are unaware of potential Generative AI (GenAI) use, but negative when GenAI use is explicitly disclosed.

Methods: A pre-registered large-scale online experiment leveraged Prolific participants to assess social impressions in diverse communication contexts, with varying levels of sender disclosure regarding GenAI use.

Key Findings: The influence of known or uncertain GenAI use on recipients' social impressions of message senders across different personal and professional contexts.

Limitations: Findings may not generalize beyond controlled experimental settings, and real-world contexts may involve more varied factors influencing perceptions.

Institution: University of Michigan

Research Area: Social Psychology, Human-AI Interaction, Generative AI Impact on Social Perception

Discipline: Social Science, Social Psychology, Human-Computer Interaction (HCI)

Sample Size: 647 participants