The Effect of Artificial Intelligence Feedback and Social Norms on Performance
Katherine Wu
21/07/2026
As artificial intelligence (AI) systems become common sources of feedback in workplaces and classrooms, understanding how people respond to AI-generated evaluations is increasingly important, particularly whether recipients distrust feedback simply because it comes from a machine.
Using randomized controlled experiments, this study examines how two factors simultaneously shape human performance: the attributed source of feedback (from AI versus from a human) and social norm information, namely information about peers' performance.
The findings suggest that social norm information raises effort across all settings, whether feedback is attributed to an AI or to a human, whereas the source of information alone has little effect. Social comparison, however, reveals how recipients respond to negative feedback. While negative feedback incentivizes greater effort if recipients are also provided with a social comparison benchmark, it does not have a statistically detectable effect in its absence. This pattern does not depend on the messenger though, when AI delivered feedback and human delivered feedback is paired with social norms, positive and negative feedback have similar effects.
Together, these results challenge the assumption that people inherently distrust AI-generated evaluations and suggest that information about relative performance affects recipients' behavior far more than the messenger's identity, demonstrating how powerful the influence of comparative performance information is in shaping motivation and productivity.