How the agent's gender influence users' evaluation of a QA system

In this paper we present the results of a pilot study investigating the effects of agents' gender-ambiguous vs. gender-marked look on the perceived interaction quality of a multimodal question answering system. Eight test subjects interacted with three system agents, each having a feminine, masculine or gender-ambiguous look. The subjects were told each agent was representing a differently configured system. In fact, they were interacting with the same system. In the end, the subjects filled in an evaluation questionnaire and participated in an in-depth qualitative interview. The results showed that the user evaluation seemed to be influenced by the agent's gender look: the system represented by the feminine agent achieved on average the highest evaluation scores. On the other hand, the system represented by the gender-ambiguous agent was systematically lower rated. This outcome might be relevant for an appropriate agent look, especially since many designers tend to develop gender-ambiguous characters for interactive interfaces to match various users' preferences. However, additional empirical evidence is needed in the future to confirm our findings.