This paper presents a comprehensive design framework for AI chatbots supporting the psychological stability of hospitalized patients. We introduce five novel contributions: (1) a vedanā model that reformulates patient psychology using Buddhist Abhidhamma concepts, identifying the vedanā-to-taṇhā conversion as the precise intervention point for AI chatbots; (2) a text-based somatic signal detection method that maps text input patterns to autonomic nervous system states via Polyvagal Theory (Porges, 2025); (3) the RET (Reception→Empathy→Tentative suggestion) protocol, translating Rogers's (1957) person-centered therapy conditions into implementable system prompt design principles; (4) a Three-Fetters prompt architecture that structurally eliminates AI's three major risks—sycophancy, hallucination, and robotic responses—by mapping them to the Buddhist concept of three fetters; and (5) an explicit AI/human responsibility boundary matrix. The framework is grounded in the Therabot RCT (Heinz et al., NEJM AI, 2025; N=210; MDD symptom reduction 51%, d=0.845–0.903). Complete Python implementation provided under MIT License.
Takeuchi et al. (Sun,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: