Researchers have developed RECAP, a novel framework designed to enhance the emotional intelligence and transparency of large language models in medical dialogue systems. This inference-time approach, based on cognitive appraisal theory, decomposes patient input into auditable stages without requiring model retraining. Evaluations across various models showed RECAP improves alignment with human judgments, particularly for smaller models, and revealed that models tend to underweight relational factors. In blinded trials, oncology fellows rated RECAP-enhanced responses significantly higher than baselines, indicating its potential for improving clinical trust in AI. AI
IMPACT Enhances emotional alignment and transparency in medical AI, potentially increasing clinical trust and adoption.
RANK_REASON The cluster contains a research paper detailing a new framework for improving AI in medical dialogue systems. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
