An Exploration of Implicit Attitudes Towards Robots Using Implicit Measures

Given the importance of setting accurate expectations of robot capabilities in humans, we explore how people form implicit competence judgments toward a robot over a prolonged interaction, and how durable those implicit judgments are. We created a live interaction between a robot deli cashier and a human customer. The robot displays a competent behavior followed by: another competent behavior (Condition A), an incompetent behavior (Condition B), or an incompetent behavior with a warning (Condition C). We measured implicit judgments of competence toward the robot using the affect misattribution procedure over time and by condition. We then measured the durability of these implicit impressions after several months. Contrary previous work, results show that there is no immediate effect of warning on guarding against a drop in implicit competence judgments. We find that the effect of warning emerge only months later.