JMIR Publications

ABSTRACT

Background:

Virtual reality exposure therapy (VRET) has emerged as an effective treatment for individuals with specific phobias, particularly for those unable to access traditional exposure therapy. However, for VRET to be fully integrated into clinical practice or the commercial market, several aspects of VRET applications still need further exploration. One such aspect is the impact of embodied conversational agents (ECAs) as virtual therapists on patient experience and treatment effectiveness.

Objective:

This study aims to investigate how the realism level and species of a virtual therapist influence user experience and perceived therapist effectiveness in a VRET intervention for dog phobia.

Methods:

We conducted a mixed-methods study where participants interacted with four different virtual character types representing virtual therapists during simulated VRET interventions for dog phobia. The four virtual character types included a realistic adult woman, a stylized adult woman, a stylized cat, and a stylized robot. We collected (1) quantitative data through standardized questionnaires assessing user experience, presence, cybersickness, and perceived therapist effectiveness, and (2) qualitative data via an open-ended preference question at the end of the study.

Results:

A total of 40 participants completed the experiment. The realistic human scored significantly lower in hedonic quality than the stylized human (z = 2.53, p = .011, padj = .015), the robot (z = 2.726, p = .006, padj = .012), and the cat (z = 3.772, p < .001, padj = .004). Additionally, the cat scored higher in hedonic quality than the stylized human (z = 2.528, p = .015, padj = .015). While no significant differences were found among the virtual character types regarding trust or willingness to share information, the therapeutic relationship skills of the robot were rated significantly lower than those of the stylized human (z = -2.315, p = .021, padj = .021). However, when asked to indicate a preferred virtual therapist, 30% of participants chose the robot, 27.5% the stylized human, 25% the cat, and 17.5% the realistic human, making the latter the least favored preferred option.

Conclusions:

The realistic human was the least suitable virtual therapist, likely due to the Uncanny Valley effect, which suggests a preference for stylized over realistic human characters when working with low-budget models. The cat's higher hedonic quality, and the robot's popularity highlight the importance of tailoring virtual therapist designs to clinical contexts and patient demographics. These findings underscore the need for further research into optimizing virtual therapist design for VRET interventions.