Abstract
STUDY OBJECTIVE: Artificial intelligence (AI) is being increasingly applied in medical education, yet its role in developing complex communication skills, such as breaking bad news, is less well defined. The Joint Royal Colleges of Physicians Training Board (JRCPTB) explored whether AI-generated guidance could complement or enhance traditional human-facilitated teaching methods in this sensitive domain. DESIGN: Two sets of guidance on breaking bad news were developed: one using a generative AI chatbot (ChatGPT) and the other through facilitated discussions among trainers and resident doctors. SETTING AND PARTICIPANTS: Each approach generated a series of summarised statements that were presented to members of the JRCPTB Specialty Advisory Committees (SACs) via an anonymous online consultation. MAIN OUTCOME MEASURES: Respondents rated their agreement with each statement using a five-point Likert scale and provided qualitative feedback. Demographic data were collected to assess variations in preferences. RESULTS: A total of 80 assessments were completed for the traditional approach and 75 for the AI approach, involving 69 doctors across 19 specialties and 11 lay representatives. Both approaches produced a common core of 11 statements with high agreement (>85%) alongside unique statements specific to each method. Overall, 61% of respondents preferred the AI-generated content, 22% preferred the traditional approach and 17% expressed no preference. Female respondents showed a statistically significant preference for the AI approach (p = 0.003) although small study numbers restrict generalisability. No differences were found based on age, ethnicity or training background. CONCLUSIONS: AI-generated content, when appropriately curated, can effectively support the teaching of sensitive communication skills, complementing traditional reflective learning methods. A hybrid model that integrates AI with human-facilitated discussions may offer a comprehensive and efficient approach to postgraduate medical education. Further research is warranted to ensure content quality, cultural and setting appropriateness, and to preserve trust in the supervisor-learner relationship.