Reducing Polarization through Generative AI

Main Researcher: Dr. Antino Kim

Overview:
This study focuses on how the timing of AI errors affects user trust, revealing that early mistakes have a greater impact on user confidence than those occurring later. The research also explores how giving users control over their reliance on AI can increase satisfaction and encourage further usage.

Key Insights:
Users are more likely to lose trust in an AI system if it makes errors early in an interaction, even if it performs accurately afterward. Conversely, users are generally forgiving of errors that occur later in a session. Additionally, when users can adjust how much they depend on AI recommendations, they report higher satisfaction and a greater willingness to rely on the AI, emphasizing the importance of user autonomy.