Popular discourses are thick with narratives of AI’s problematic functions and outcomes, yet there is little understanding of how everyday people (i.e., non-experts) consider AI activities to constitute bad behavior. This study explores folk theories of large-language models’ bad behavior through inductive analysis of interviews with non-experts (N = 28). Results indicate perception of AI bad behavior at low construal levels is limited to high-level notions of acts against humans and acts against reality. However, more discrete and varied notions of AI behaving badly manifest when evaluating specific AI behaviors. Types of observed behaviors considered bad mostly align with moral foundations. A novel characterization emerged across all observed behaviors: Some variations on nonperformance and social discordance. This characterization is interpreted to suggest that discordant nonperformance could be the essence of AI bad behavior as a folk category.
No takes yet. Share an insight, caveat, or question.
Jaime Banks (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: