Empirical Investigations into the Believability of Robot Affect

The concept of “believability” (of an agent) is difficult to pin down, making its value questionable for experimental design, quantitative empirical evaluations, and explanations of people’s perceptions of agents in general. We propose to replace “believability” with a set of finer-grained notions based on people’s attitudes that are better suited to these uses. Based on our analysis, we demonstrate an experimental methodology to evaluate subjects’ attitudes toward robot affective states, one which allows us to get at various aspects of believability that would be difficult to achieve with more coarse-grained notions of believability.