Adaptivity and Revealed Robot Aversion in Human-Robot Collaboration: A Field-in-the-Lab Experiment

We study human-robot collaboration in a controlled experiment run in a realistic production environment. Participants completed a sequential task in pairs, where one worker (Worker 1) decided whether to pass intermediate components to a coworker or not. Depending on the treatment, the coworker was either another human participant or a physical industrial robot. The coworker-setup was either static or adaptive, with adaptive coworkers' productivity being influenced by Worker 1's performance in the task. We find strong evidence of robot aversion: workers were significantly less likely to pass intermediate products to their coworkers in the robotic as compared to the human treatments. This was despite overall productivity was identical across treatments. In a subsequent responsibility attribution task, participants also attributed greater responsibility to the robots, indicating a systematic bias in social evaluation of machine coworkers. Adaptivity only marginally affected these outcomes. Our results demonstrate that cooperation and responsibility attribution in hybrid teams depend not only on performance but also on social perceptions of artificial agents, highlighting behavioural frictions that may constrain the effective integration of robots into human work environments.