To be healed or to be hacked

The high prevalence of mental health disorders worldwide has created a big demand on mental health care professionals. To relieve personnel and provide increased access to care, the use of AI and VR has been researched for many decades. As the human-centered implementation of AI remains scarce, we first assessed mental health patients’ concerns about the use of AI for therapy in an online questionnaire. The pre-study revealed the importance of data autonomy and transparency, which was explored in a subsequent in-person VR study regarding the effect on self-disclosure and trust towards the embodied AI. Results suggest a higher level of self-disclosure with higher data autonomy. Transparency did not influence level of self-disclosure and both autonomy and transparency showed no effect on trust towards the AI. Additionally, an effect of gender on trust was found, showing higher levels of trust for women compared to men. This work contributes to the ethical discussion around AI in mental health care with a human-centered AI perspective and insights into patients’ concerns on this development.