Once One Fails, All Are Suspect: Understanding Error Generalization in AI

Authors: L Dai, Z Wang, L Chen, J Jin

Published: 2026

Publication: 2026•scholarspace.manoa.hawaii.edu

AI errors lead to broader negative generalizations about other AI systems compared to human errors, largely due to perceptions of AI's inflexibility and inability to learn from mistakes.

Methods: Conducted four one-factor experiments across distinct contexts to compare human responses to AI errors and human errors.

Key Findings: Generalization of error perceptions from one AI system to others, and psychological mechanisms driving this process.

Limitations: Exact contexts and diversity of experimental settings or participant demographics are not detailed.

Institution: Shanghai International Studies University

Research Area: Socio-Economic Impacts of AI, Algorithmic Systems

Discipline: Computer Science, Artificial Intelligence