Reproduce “Towards Explainable and Safe Conversational Agents for Mental Health: A Survey” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Surjodeep Sarkar, Manas Gaur, L. Chen, Muskan Garg, Biplav Srivastava, Bhaktee Dongaonkar
This survey critically evaluates current conversational agents for mental health, highlighting their limitations in explainability, safety, and clinical grounding. It proposes a framework for building knowledge-infused, trustworthy virtual mental health assistants and advocates for new evaluation metrics for responsible development.
Key insights
- Current VMHAs fail to engage in reflective, context-aware, and clinically grounded conversations, providing superficial support.
- The 'black box' nature of LLMs introduces safety risks in mental health, including hallucinations and harmful advice, due to a lack of explainability.
- Integrating machine-readable knowledge is vital for developing safe, explainable, and contextually aware VMHAs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.