Role-Playing with Virtual Actors in an Online Communication Training: The Influence of Different Character Representations on User Experience and Learning Outcome

Abstract

Effective communication is an important workplace requirement and can be improved through soft skill training programmes. A frequently used method for training interpersonal skills is through role-play. Faculty of Skills, the training agency this research is conducted for, uses video role-play in their online training programmes. These videos are currently created by recording actors in a studio, however, this is an expensive and time-consuming process. Therefore, this research investigates whether digital alternatives (i.e., deepfakes and virtual humans) could replace actors in these videos without coming at the expense of the learning outcome and user experience. Thus, the following research question was formulated: ”How do different character representations in a virtual communication training influence the experience and learning outcome of trainees?”. To answer this, a between-subjects study was designed, in which participants (N = 103) were randomly assigned to either the A) actor, B) deepfake, or C) virtual human version. Participants completed a short training programme and filled out a survey afterwards. The results show that there is no significant difference in learning outcome between the three groups. Additionally, participants did not rate the quality of the training significantly different. The quality of the characters, however, was rated significantly lower by participants from the virtual human group. Yet this version was rated as more inventive than the other two. No significant differences where found for other components of the user experience, such as how enjoyable, interesting, or attractive the training was. Furthermore, no significant differences were found in the feeling of engagement, presence or immersion. Finally, four video characters (2 male, 2 female) of each version were rated based on several statements. The virtual humans were generally rated worse than the other two groups. An explanation for this could be due to the less natural sounding text-to-speech (TTS) voice, or perhaps due to factors such as the length of the video or design choices regarding the appearance of the character. On the other hand, gender of the character did not seem to influence the perception. Based on the results, Faculty of Skills could potentially start tests with actual clients, in which they offer them personalised role-plays that are delivered faster and at a lower price. This way, it can also be verified whether these results hold up after a full-length training programme as well.