The Influence of Robot Autonomy on Sense of Control and Trust Towards a Robot

Autonomous robots can make decisions that lead to outcomes that are better, equal to, or worse than those requested by a human. However, how the value of these outcomes shapes perceptions of robot autonomy remains underexplored. We conducted four online experiments examining how different forms of robot decision-making autonomy influence human–robot interaction. Participants played a game in which they selected a box for a robot avatar to collect, with each box yielding a specific number of points. Some robots were described as autonomous and could disobey instructions. We compared robots that disobeyed to produce a better outcome for the participant (helpful autonomy) with robots that disobeyed but achieved the same outcome (same-outcome autonomy). Experiments 3 and 4 further distinguished helpful autonomy aimed at avoiding losses from autonomy aimed at obtaining additional gains. Trust was assessed using a post-task trust game. Results showed that helpful autonomy generally increased trust compared to same-outcome autonomy, except when loss avoidance was unreliable. Trust was unrelated to perceived control. Perceived control was higher for same-outcome autonomy in Experiments 1–2, where non-autonomous and random robots were present, but higher for helpful autonomy in Experiments 3–4, indicating that perceived control depends on both robot autonomy and the behavioral context in which it is experienced. Finally, helpful autonomy produced bimodal distributions in ratings of robot cooperativeness and experienced sense of joint agency, suggesting that such behavior can elicit qualitatively different attitudes toward the robot.