Beyond Human vs. Machine: An Interactionist Model of AI Evaluation

Moving beyond simple comparisons between AI and human evaluators, this study investigates how users respond to feedback as a function of evaluator type (AI vs. human), feedback valence (positive vs. negative), and authority framing (high vs. low). Drawing on Psychological Distance Theory and Expectation Violation Theory, we propose an interactionist model in which feedback effectiveness depends on the joint configuration of these factors. Using a 2 × 2 × 2 between-subjects experiment with 353 participants, we examined the effects of evaluator type, feedback valence, and authority cues on perceived productivity and word-of-mouth intentions. Findings highlight that AI-mediated evaluation operates in a conditional rather than uniform. AI performs best as a credible source of data-driven positive reinforcement when authority is clearly established, but is vulnerable to resistance when delivering negative feedback without sufficient legitimacy cues. The results support hybrid evaluation systems that leverage AI and human strengths in complementary ways rather than positioning AI.