PUB - Publikationen an der Universität Bielefeld

In cooperative settings the success of the team is interlinked with the performance of the individual members. Thus, the possibility to address problems and mistakes of team members needs to be given. A common means in human-human interaction is the attribution of blame. Yet, it is not clear how blame attributions affect cooperation between humans and intelligent virtual agents and the overall perception of the agent. In order to take a first step in answering these questions, a study on cooperative human-agent interaction was conducted. The study was designed to investigate the effects of two different blaming strategies used by the agent in response to an alleged goal achievement failure, that is, self-blame (agent blames itself) followed by an apology versus other-blame (agent blames the user). The results indicate that the combination of blame and trust repair enables a successful continuation of the cooperation without loss of trust and likeability.