The field of AI-powered mental health support is moving towards a more nuanced understanding of the values and harms associated with the use of large language models (LLMs) in this context. Researchers are investigating the potential benefits and risks of LLMs, such as ChatGPT, in providing support for individuals with mental health concerns. The focus is on developing a deeper understanding of how LLMs can be designed to provide effective and safe support, while minimizing the risks of harm. This includes exploring the role of transparency, bias, and privacy in LLMs, as well as the importance of human oversight and critical validation. Noteworthy papers in this area include:

• A study on AI chatbots for mental health, which identified key values such as informational support, emotional support, and privacy, and provided design recommendations for minimizing risks.
• An analysis of Reddit posts and comments, which found that users value ChatGPT as a safe and non-judgmental space for discussing mental health concerns, but also raised concerns about the potential risks of incorrect health advice and privacy concerns.