The power of justifications to repair human-robot trust, even under moral disagreement

摘要

To avert criticism and losses of trust, robots that adopt social roles in the near future will have to be aware of and follow the norms of the communities in which they operate. However, norms often conflict with one another, and resolving such conflicts requires prioritizing one norm and violating the other, conflicting norm. As a result, robots will face moral disapproval from at least some of their human interaction partners. We investigate a powerful tool that humans use-and autonomous agents should use-to manage such moral disapproval and maintain trust: justifications, which explain not just why the agent acted but what norms and values the action upheld. In three experiments (N = 3,596), we demonstrate, replicate, and generalize that justifications, more than mere explanations, mitigate moral disapproval and recover robots' perceived trustworthiness, even when the robot's action is in direct moral disagreement with the human observer. We conclude that people simultaneously blame the robot for its specific norm-violating action and appreciate the robot's integrity to make trustworthy decisions. Justifying norm-violating actions may allow robots to become better integrated into human communities and adopt social roles that will involve morally significant decisions.