NoRobot's perfect : trust repair in the face of agent error. How do individual factors influence trust development in human-agent teams?

People increasingly work together with autonomous agents. To work successfully, the human needs to possess an appropriate level of trust towards the autonomous agent, not too much and not too little. Since trust may decrease in the inevitable event of an automation error, effective trust repair strategies are required to restore appropriate trust levels. This is crucial to ensure that the human-agent team (HAT) keeps on working most productively. The present study aimed to investigate how explanation and agent type relate to a successful trust restoration in a HAT. Also, the effect of individual factors such as propensity to trust, forgiveness, and perceived threat on trust development were experimentally explored. Participants (N=38) completed two missions in a virtual reality house search with an autonomous agent that served as an advisor for the safety of the environment. During the mission, humans' trust in the autonomous agent was purposefully violated to enable examination of trust development in the occurrence of errors. In one of the missions, the agent attempted trust repair by providing an explanation for the error. Although no significant effects were found in this investigation of preliminary data, visual inspections revealed some promising insights that will be discussed.