Researchers at the University of Limerick conducted a study involving 189 knowledge workers to evaluate human interaction with AI agents. Participants collaborated with three types of virtual colleagues: a text-based interface, a small robot avatar, and human-like avatars.
The human-like avatars were represented by two versions: a female named Johanna and a male named Johan. Participants were asked to rate these agents based on trust and credibility, and to allocate future development resources to them as a form of compensation.
The results indicated that human-like agents received higher trust and credibility scores compared to robot-like versions. While there was no significant difference in trust between Johanna and Johan, the female agent received statistically lower compensation than her male counterpart.
Participants generally reported that they did not believe the gender of the AI agent influenced their feedback. Additionally, the study noted that the text interface and the robot avatar performed similarly in terms of trust and credibility ratings.
