Integrating AI in Psychotherapy: An Investigation of Trust in Voicebot Therapists

The use of Artificial Intelligence (AI) applications, such as voicebots, holds potential to alleviate the current shortage in mental health services. However, it is unclear whether clients would place the same level of trust in these agents as they do in human therapists. This study aims to explore clients' trust in self-disclosure and privacy when interacting with a voicebot therapist. Additionally, it investigates the potential impact of incorporating explainable AI (XAI) on trust perceptions. In a pre-registered, between-subjects design (2x2: human vs. voicebot, explanation vs. no explanation) study, 564 participants were presented with a pre-recorded therapy simulation. Participants were asked whether they would trust the human or voicebot therapist and whether they would be willing to use it in the future. The results revealed no significant difference in trust between the voicebot and the human therapist. Furthermore, increased transparency through XAI enhanced trust in the voicebot's overall functionality and led to higher levels of trust in its privacy measures.