DataSample Avatar

The final sample consisted of adult participants who were either currently engaged in psychological therapy or had received psychological treatment within the previous six months. This inclusion criterion was adopted to ensure that respondents had recent and relevant experience with therapeutic contexts, thereby increasing the ecological validity of the avatar selection task.

Participants were recruited using a convenience sampling strategy and completed an online survey administered via a mobile-optimized platform. After providing informed consent, respondents were exposed to a decision task in which they were asked to select a preferred therapist avatar from a set of AI-generated faces. The avatars varied systematically along two primary dimensions—race (e.g., Black and White) and gender (male and female)—as well as in visual realism (photorealistic vs. stylized representations), depending on the experimental condition.

Each participant viewed a subset of avatars presented individually on their device. The faces were explicitly described as illustrative representations only, with no differences in underlying behavior or competence attributed to any avatar. Participants were made aware of this aspect to isolate the effect of visual identity cues on selection decisions. The composition of the avatar set was randomized across respondents, such that each participant encountered a unique combination of faces drawn from a larger pool, thereby minimizing potential image-specific biases.