Advances in AI, including large language models, enable virtual health assistants (VHAs) to tailor messaging to cultural and individual needs.However, to fully realize this potential, VHA voices must match or exceed human voices in influencing behavioral intentions.We conducted a multi-study investigation on whether matching text-to-speech (TTS) voices to human voices across vocal traits could promote colorectal cancer (CRC) screening among Black or African American adults.Study 1 (N = 313) tested whether a TTS voice prosodically matched to a human voice could elicit equivalent screening intentions using a White male VHA.It also evaluated whether increasing vocal confidence via SSML would enhance effectiveness.A user preference study (N = 317) assessed VHA identity and delivery modality preferences, finding strong support for a Black female VHA.Building on these results, Study 2 (N = 106) replicated the TTS-human voice comparison using a Black female VHA.Across studies, matching TTS voices to human voices on frequency, speech rate, and pitch variation achieved equivalent * Both authors contributed equally to this research.

Figure 1: The virtual health assistant with example dialogue used in the stress-coping gratitude journaling intervention

Virtual health assistants, digital characters designed to guide patients through healthcare interventions, offer scalable solutions, but their effectiveness may depend on user perceptions of realism. This study investigated how the visual fidelity and voice modality of virtual health assistants influence perceived realism, perceived voice naturalness, and intentions to screen for colorectal cancer among Black American adults aged 45–75. In a 2 × 4 factorial between-participants experiment (N = 266), participants were randomized to virtual health assistants varying in visual fidelity and voice modality. Structural equation modeling revealed that voice naturalness had a strong positive effect on perceived realism, which in turn significantly predicted screening intentions. Visual fidelity also contributed directly to perceived realism, although to a lesser extent. These results suggest that enhancing voice naturalness in virtual health assistants may be a critical design priority for improving engagement and promoting preventive health behaviors in digital health interventions.

Virtual humans are employed in various contexts, including mental health interventions, to encourage users to adopt healthy behaviors. Establishing rapport with users can enhance the effectiveness of these virtual agents. One way to build rapport is by matching the personality between the virtual human and the participant, which may elicit a similarity-attraction effect, shown to increase trust and likeability in interactions. Despite the potential of computer-generated voices to convey personality through the manipulation of vocalic properties, prior research has primarily focused on non-vocal aspects of personality. To address this gap, we conducted an online study that altered a virtual human's vocalic properties to represent high or low extroversion, with a focus on the role of rapport in promoting mental wellness. In this study, a virtual human provided information on stress-reducing mental-wellness practices to 165 participants. Our findings suggest that synthesizing vocalic properties to resemble a low extroversion level can enhance the persuasiveness of virtual humans and improve rapport, as indicated by participants' self-reported intention to engage in mental-wellness practices.

Virtual humans demonstrate the ability to act as non-judgmental conversational partners, eliciting greater self-disclosure. However, it is unclear what virtual human and conversational characteristics are important when self-disclosing. To address this gap, we conducted a set of qualitative, semi-formal interviews (n = 17) among computer science students to investigate participant mental models of willingness to disclose to virtual humans and characteristics of virtual humans that affect their self-disclosure. Our findings indicate that participants' mental models of virtual humans are largely inconsistent with current literature. This inconsistency appears to eliciting hesitancy and discomfort with virtual humans. Furthermore, trust and listening were identified as two primary characteristics of a virtual human interaction that are valuable towards willingness to disclose. Additionally, these characteristics were also valued in different ways for virtual humans in comparison to real humans. From the interviews, we identify and provide guidelines of designing virtual human interactions and conversations to elicit greater willingness to disclose.

Learning how to program is perceived by many college students as difficult. Factors that influence novices’ success in computer programming learning include the student’s mental model of programming, computer playfulness during training, and programming anxiety. Programming anxiety (PA) is a psychological state engendered when a student experiences or expects to lose self-esteem in confronting a programming task. Students’ achievements have been seen to be negatively affected by programming anxiety. Others have investigated storytelling and tangible block programming to facilitate learning programming skills, but such efforts did not focus on analyzing their effects on users’ anxiety toward programming. In this work, we present fableBlocks, a tangible block-based programming (BBP) environment that relies on storytelling to mitigate programming anxiety. We adapted an existing BBP environment to incorporate storytelling. In a comparative pilot study, fableBlocks outdid a conventional GUI-based environment in usability and user-experience scores. Although not statistically significant, participants demonstrated lower levels of PA with fableBlocks.