| dc.description.abstract | A Dutch tech company, Ordina, was developing a VR training application for job interviewees. By order of that company, we were firstly requested to find the best medium to present the application, a head-mounted display (HMD) or a Desktop monitor. The best medium, in this case, elicits the highest perceived immersion. To find the best medium we conducted a job interview under three different conditions during the development of the application. The three conditions were the Desktop condition, the HMD condition, and the real life control condition. The first factor that differed across all conditions was the presentation method. To measure the perceived immersion of the participants we used the Slater-Usoh-Steed Questionnaire and the Presence Questionnaire. We expected that the Real condition had the highest perceived immersion, followed by the HMD condition with the Desktop condition being the lowest. In contrast to our expectation, the results from the questionnaires showed that participants had an equally high perceived immersion in the Real and HMD condition. However, as expected, it also revealed that the Desktop condition had the lowest perceived immersion. Secondly, the company requested that we create a conversational partner who was able to elicit human-like verbal behavior in participants. To see if our conversational partner could achieve this, we analyzed the verbal behavior from the job interviews mentioned above. The second factor that differed was the conversational partner. This partner was an embodied conversational agent in the HMD and Desktop version or a real human in the real version. To measure the verbal behavior of the participants, we used the participants’ total words spoken, words longer than six letters, disfluencies, and types. We expected the highest scores on all verbal behavior variables for the real interview. Besides, we expected the HMD and Desktop to have the same scores on the verbal behavior variables. The results from the verbal behavior variables showed that the Real condition indeed had the highest scores. As expected, the results also showed that the Desktop and HMD versions had the same scores. We advise investing most resources in developing the conversational agent with the primary focus on natural language processing. The reason for this is that the verbal behavior variables indicated that participants in both the HMD and Desktop version did not behave similarly to the Real condition. Participants' behavior might resemble real conversations more by developing the natural language processing of the application. | |
