Being Sorry is the Hardest Thing: How Robots can Apologize and Learn from Mistakes to Restore People's Trust

摘要

Robots that move around people in the workplace will make mistakes, such as getting too close to someone or blocking the way. These mistakes negatively impact people’s experience with and attitudes toward robots. For a robot to successfully recover from mistakes during navigation, we need deeper insights into the effects of these mistakes and the effectiveness of different recovery strategies. We conducted an online survey (N=219) to assess people’s trust in a robot before and after errors, using three recovery strategies (communicating learning capability, apologizing, or a combination of both). Results show that trust dropped significantly after a navigational error but could be effectively restored, especially when the robot apologizes and communicates its ability to learn from it. When the robot does not try to recover from a mistake, the effects are very detrimental. These results highlight the importance of appropriate error recovery strategies for robots to enhance human-robot interactions.