Inferring capabilities of intelligent agents from their external traits

We investigate the usability of human-like agent-based interfaces. In an experiment with a traveladvisory system, we manipulate the "human-likeness" of the agent interface. We demonstratethat users of the more human-like agents form an anthropomorphic use image of the system: theyact human-like towards the system and try to exploit typical human-like capabilities they believethe system possesses. Unfortunately, this severely reduces the usability of systems that lookhuman but lack human-like capabilities (overestimation effect). Furthermore, we demonstratethat the use image users form of an agent-based system is inherently integrated (as opposed tothe compositional use image they form of conventional interfaces): feedforward cues providedby the system do not instill user responses in a one-to-one matter, but are instead integrated intoa single use image. Consequently, users try to exploit capabilities that were not signaled by thesystem to begin with, thereby further exacerbating the overestimation effect.