Examining Trust and Reliance in Collaborations between Humans and Automated Agents

Abstract

Human trust and reliance in artificial agents is critical to effective collaboration in mixed human computer teams. Understanding the conditions under which humans trust and rely upon automated agent recommendations is important as trust is one of the mechanisms that allow people to interact effectively with a variety of teammates. We conducted exploratory research to investigate how personality characteristics and uncertainty conditions affect human-machine interactions. Participants were asked to determine if two images depicted the same or different people, while simultaneously considering the recommendation of an automated agent. Results of this effort demonstrated a correlation between judgements of agent expertise and user trust. In addition, we found that in conditions of high and low uncertainty, the decision outcomes of participants moved significantly in the direction of the agent’s recommendation. Differences in reported trust in the agent were observed in individuals with low and high levels of extraversion.