Impact of Facial Self-Similarity and Gender of a Storytelling Virtual Character

Sammanfattning: Technical advancements allow for embodied virtual agents to not only be increasingly human-like, but also to behave and look like particular individuals. As biases towards self-similarity have been found in human-human studies, it is of interest to explore to what extent this applies to virtual characters (VCs). This work set out to extend on previous research that has investigated the effects of facial self-similarity in VCs, and explore it in the context of empathic emotion. For this aim, a method for creating facially similar virtual characters was developed and a user study conducted where 13 participants were told autobiographical stories by a virtual character that either did or did not resemble them facially and/or in gender category. The participants' first impressions and emotional responses were measured. The results showed that even though similarity was not explicitly perceived, a bias might exist towards more positive impressions of self-similar characters, especially in terms of gender category. Regarding the emotional responses, the results did not allow for discovering any difference between conditions but pointed to some interesting differences in comparison to what was hypothesized. The immense ways in which the appearances of virtual characters can be altered provides possibilities to influence the interaction with them. However, although biases might exist on a general level, it is difficult to predict the human responses in individual cases. Virtual characters might make possible a more human-like interaction with technology, however, it might also mean that our reactions to them are influenced by more parameters and our relations to them become even more like those with other humans: complex.