Presentation
CANCELED - Emotional Engagement and Alliance of a GenAI-VR System for Mental Health Support: An Interview Study
DescriptionBackground
With the increasing prevalence of mental health issues and a shortage of clinicians, integrating generative artificial intelligence (GenAI) with virtual reality (VR) holds significant promise. However, the application of current VR systems in mental health remains limited due to their reliance on standardized rather than personalized content. Studies have shown that personalization, such as customizable avatars and environments, can enhance user experience and engagement, which supports symptom reduction. At the same time, psychotherapy research has demonstrated that engagement cannot be activated in isolation; it depends on enabling conditions such as feasibility, acceptability, safety, and fit, which map directly onto core implementation outcomes. These observations emphasize that, to achieve therapeutic effectiveness, the design of VR therapy systems must simultaneously optimize engagement and implementation outcomes. This study proposes developing a GenAI-VR system that will provide personalized and meaningful mental support by utilizing immersive, customizable VR environments and avatars, thereby fostering a more profound sense of social connectedness and emotional engagement compared to existing VR-based mental health tools. To guide this process, we focus on Norman’s three levels of design to address emotional engagement, and Proctor’s implementation outcomes framework to assess feasibility, acceptability, and fit of the application within clinical contexts. Norman’s three levels of design, including visceral, behavioral, and reflective, provide a structured framework for understanding how users emotionally engage with technology. Visceral design addresses immediate sensory appeal, behavioral design focuses on usability and functional flow, and reflective design captures personal meaning and value. Proctor’s implementation outcomes framework provides a widely recognized taxonomy for evaluating how interventions can be adopted and sustained in real-world settings. From this framework, we focus on acceptability, safety, and fit. We prioritize these outcomes because they represent enabling conditions for engagement and alliance in psychotherapy research.
Objectives
This study focuses on developing a prototype of the proposed GenAI-VR system designed to deliver personalized mental health support. User interfaces were designed, developed, and evaluated with target users through semi-structured interviews. The objectives were to assess the users' emotional engagement and alliance with the GenAI-VR system within psychotherapy settings.
Experimental Procedure
The study began with the development of the prototype, which included designing the user interfaces, user and therapist avatars, and therapy environments. Twenty-five user avatars, including both human and animal, five therapist avatars, and nineteen therapy environments, including indoor, natural, and fantasy settings, were designed to facilitate virtual engagement with the therapy sessions.
The interfaces were printed and trimmed to construct the paper prototype. To evaluate the prototype, we invited ten individuals experiencing mild to moderate depression. Upon arrival, the GenAI-VR system’s features and the experimental procedure were explained to the participants. Next, the participants performed interactive tasks, including navigating through the interfaces and selecting a therapy session by choosing their avatars, therapist, and environments using the paper interfaces. Participants were then asked to imagine themselves as their chosen avatar, with the researcher embodying the selected therapist avatar, and visualize being in their chosen environment. Next, a semi-structured interview session was conducted between the researcher and the participants, which included discussions about the participants' experiences and opinions regarding the usability of the GenAI therapist prototype. Finally, participants rated the prototype's usability using the Usability, Satisfaction, and Ease of Use (USE) questionnaire, which had four dimensions: usefulness, ease of use, ease of learning, and satisfaction. Additionally, the interview sessions were transcribed and analyzed using a thematic analysis approach.
Results
The thematic analysis revealed one global theme: the potential of GenAI-VR systems in mental healthcare, as well as several basic themes. Participants expressed positive views towards the prototype. The prototype's environments were praised for their coziness, friendliness, and calming elements. Participants appreciated the use of avatars within the platform, finding them helpful for creating a comfortable and less intimidating therapy environment. The usability of the prototype was reported to be professional and easy to navigate, with appreciation for its user-friendly interface and design aesthetics. Participants emphasized the incorporation of human-like problem-solving approaches based on real-life experiences, integration of specialized knowledge from diverse backgrounds, and continuous learning through regular updates and feedback, to ensure that GenAI can address mental health needs through personalized interactions. They noted that the non-judgmental nature of AI, combined with anonymity, can help build trust and encourage open communication. The convenience and accessibility of the GenAI therapist, which will enable users to seek help at any time, even when human therapists are unavailable, as well as the ease of access through smartphones and cost-effectiveness, were also highlighted. However, participants expressed concerns about GenAI's ability to replicate human emotions, as well as the possible leakage of personal information. They provided ideas on GenAI and human therapists' collaboration, where AI could provide initial support and filter essential information from patients' inputs, allowing human therapists to focus on key details for treatment. Finally, participants emphasized that the GenAI-VR system can enhance mental health awareness by making it accessible, less intimidating, and motivating a broader audience, thereby reducing stigma and encouraging early intervention.
The results from the USE questionnaire further validated participants' acceptance of the GenAI application prototype. Participants reported a high level of acceptance in terms of usefulness, ease of use, ease of learning, and satisfaction. The usefulness dimension indicates a belief that the application would have a positive impact on their mental health. The ease-of-use dimension suggests that participants found the application effortless to use, and the ease of learning component demonstrates that participants needed minimal effort to learn how to use the application. And the high satisfaction levels suggest that the application has high usability, supporting its potential benefits in aiding users' mental health.
Discussion and Conclusion
The findings of the experiment reflected each of Norman's three levels of design. For example, the visceral aspect of the design was addressed through avatars and environments that felt inviting to the participants. The admiration for the application's user-friendly interface, professional and intuitive navigation through the modules, and the high scores on the usefulness dimension of the USE questionnaire reflect the behavioral level of the design. The reflective aspect of the prototype was evident in participants' sense of accomplishment, their perception of the therapy environment as less intimidating, the personalized and non-judgmental nature of AI, which fosters trust and encourages open communication, and the high satisfaction scores on the USE questionnaire. Collectively, these levels indicate how the GenAI therapist application can elicit emotional engagement, which is essential for sustained therapeutic participation.
The results supported Proctor's implementation outcomes framework. Acceptability was supported by the perceived suitability of the system and the participants' positive views regarding the application. Safety was reflected in the participants' comments on the non-judgmental nature of AI, its ability to foster trust and encourage open communication, as well as the absence of adverse events or distress. Fit was demonstrated by participants' perception of GenAI as motivating a broader audience, collaborating with human therapists, and being appropriate within therapeutic contexts. These findings further indicate the feasibility, acceptability, and suitability of the GenAI-VR system within psychotherapy.
The experiment demonstrated participants' emotional engagement and alliance with the prototype, reflecting acceptability and potential for the GenAI-VR system in mental healthcare. Future studies will employ co-design and iterative testing to refine the system, ensuring its engagement and feasibility for real-world clinical implementation.
With the increasing prevalence of mental health issues and a shortage of clinicians, integrating generative artificial intelligence (GenAI) with virtual reality (VR) holds significant promise. However, the application of current VR systems in mental health remains limited due to their reliance on standardized rather than personalized content. Studies have shown that personalization, such as customizable avatars and environments, can enhance user experience and engagement, which supports symptom reduction. At the same time, psychotherapy research has demonstrated that engagement cannot be activated in isolation; it depends on enabling conditions such as feasibility, acceptability, safety, and fit, which map directly onto core implementation outcomes. These observations emphasize that, to achieve therapeutic effectiveness, the design of VR therapy systems must simultaneously optimize engagement and implementation outcomes. This study proposes developing a GenAI-VR system that will provide personalized and meaningful mental support by utilizing immersive, customizable VR environments and avatars, thereby fostering a more profound sense of social connectedness and emotional engagement compared to existing VR-based mental health tools. To guide this process, we focus on Norman’s three levels of design to address emotional engagement, and Proctor’s implementation outcomes framework to assess feasibility, acceptability, and fit of the application within clinical contexts. Norman’s three levels of design, including visceral, behavioral, and reflective, provide a structured framework for understanding how users emotionally engage with technology. Visceral design addresses immediate sensory appeal, behavioral design focuses on usability and functional flow, and reflective design captures personal meaning and value. Proctor’s implementation outcomes framework provides a widely recognized taxonomy for evaluating how interventions can be adopted and sustained in real-world settings. From this framework, we focus on acceptability, safety, and fit. We prioritize these outcomes because they represent enabling conditions for engagement and alliance in psychotherapy research.
Objectives
This study focuses on developing a prototype of the proposed GenAI-VR system designed to deliver personalized mental health support. User interfaces were designed, developed, and evaluated with target users through semi-structured interviews. The objectives were to assess the users' emotional engagement and alliance with the GenAI-VR system within psychotherapy settings.
Experimental Procedure
The study began with the development of the prototype, which included designing the user interfaces, user and therapist avatars, and therapy environments. Twenty-five user avatars, including both human and animal, five therapist avatars, and nineteen therapy environments, including indoor, natural, and fantasy settings, were designed to facilitate virtual engagement with the therapy sessions.
The interfaces were printed and trimmed to construct the paper prototype. To evaluate the prototype, we invited ten individuals experiencing mild to moderate depression. Upon arrival, the GenAI-VR system’s features and the experimental procedure were explained to the participants. Next, the participants performed interactive tasks, including navigating through the interfaces and selecting a therapy session by choosing their avatars, therapist, and environments using the paper interfaces. Participants were then asked to imagine themselves as their chosen avatar, with the researcher embodying the selected therapist avatar, and visualize being in their chosen environment. Next, a semi-structured interview session was conducted between the researcher and the participants, which included discussions about the participants' experiences and opinions regarding the usability of the GenAI therapist prototype. Finally, participants rated the prototype's usability using the Usability, Satisfaction, and Ease of Use (USE) questionnaire, which had four dimensions: usefulness, ease of use, ease of learning, and satisfaction. Additionally, the interview sessions were transcribed and analyzed using a thematic analysis approach.
Results
The thematic analysis revealed one global theme: the potential of GenAI-VR systems in mental healthcare, as well as several basic themes. Participants expressed positive views towards the prototype. The prototype's environments were praised for their coziness, friendliness, and calming elements. Participants appreciated the use of avatars within the platform, finding them helpful for creating a comfortable and less intimidating therapy environment. The usability of the prototype was reported to be professional and easy to navigate, with appreciation for its user-friendly interface and design aesthetics. Participants emphasized the incorporation of human-like problem-solving approaches based on real-life experiences, integration of specialized knowledge from diverse backgrounds, and continuous learning through regular updates and feedback, to ensure that GenAI can address mental health needs through personalized interactions. They noted that the non-judgmental nature of AI, combined with anonymity, can help build trust and encourage open communication. The convenience and accessibility of the GenAI therapist, which will enable users to seek help at any time, even when human therapists are unavailable, as well as the ease of access through smartphones and cost-effectiveness, were also highlighted. However, participants expressed concerns about GenAI's ability to replicate human emotions, as well as the possible leakage of personal information. They provided ideas on GenAI and human therapists' collaboration, where AI could provide initial support and filter essential information from patients' inputs, allowing human therapists to focus on key details for treatment. Finally, participants emphasized that the GenAI-VR system can enhance mental health awareness by making it accessible, less intimidating, and motivating a broader audience, thereby reducing stigma and encouraging early intervention.
The results from the USE questionnaire further validated participants' acceptance of the GenAI application prototype. Participants reported a high level of acceptance in terms of usefulness, ease of use, ease of learning, and satisfaction. The usefulness dimension indicates a belief that the application would have a positive impact on their mental health. The ease-of-use dimension suggests that participants found the application effortless to use, and the ease of learning component demonstrates that participants needed minimal effort to learn how to use the application. And the high satisfaction levels suggest that the application has high usability, supporting its potential benefits in aiding users' mental health.
Discussion and Conclusion
The findings of the experiment reflected each of Norman's three levels of design. For example, the visceral aspect of the design was addressed through avatars and environments that felt inviting to the participants. The admiration for the application's user-friendly interface, professional and intuitive navigation through the modules, and the high scores on the usefulness dimension of the USE questionnaire reflect the behavioral level of the design. The reflective aspect of the prototype was evident in participants' sense of accomplishment, their perception of the therapy environment as less intimidating, the personalized and non-judgmental nature of AI, which fosters trust and encourages open communication, and the high satisfaction scores on the USE questionnaire. Collectively, these levels indicate how the GenAI therapist application can elicit emotional engagement, which is essential for sustained therapeutic participation.
The results supported Proctor's implementation outcomes framework. Acceptability was supported by the perceived suitability of the system and the participants' positive views regarding the application. Safety was reflected in the participants' comments on the non-judgmental nature of AI, its ability to foster trust and encourage open communication, as well as the absence of adverse events or distress. Fit was demonstrated by participants' perception of GenAI as motivating a broader audience, collaborating with human therapists, and being appropriate within therapeutic contexts. These findings further indicate the feasibility, acceptability, and suitability of the GenAI-VR system within psychotherapy.
The experiment demonstrated participants' emotional engagement and alliance with the prototype, reflecting acceptability and potential for the GenAI-VR system in mental healthcare. Future studies will employ co-design and iterative testing to refine the system, ensuring its engagement and feasibility for real-world clinical implementation.
Event Type
Oral Presentations
TimeTuesday, March 2410:30am - 11:00am EDT
LocationNassau
Digital Health