A Qualitative Study of Human Theorizing about Robot Bodily Behavior

This research investigates the cognitive frameworks individuals use to understand and justify robot behaviour. Participants watched a robot learning how to manipulate objects through unguided experiential learning. Across multiple rounds of try-and-fail, participants were asked to predict the robot's performance with unfamiliar objects and speculate on its cognitive operations, including reasoning, sensing, and actuation. To assess the potential impact of performance on theorization, participants were divided into two groups: one exposed to a predominantly successful robot, and the other to a predominantly failing robot. The findings reveal intriguing patterns: participants, regardless of their condition, consistently engaged in theorizing about the robot in similar quantities and ways. Those who observed the robot failing more frequently often subsequently predicted success, perhaps due to a desire to offer the robot "another chance", or because probabilistically they felt success was more likely following several failures.