Mitigating Cultural and Linguistic Discordance: A Gen AI-Powered Training on Culturally Competent Clinical Communication

Abstract

Miscommunication is way too common in culturally and linguistically discordant clinical communication contexts. Cultural competence training provides healthcare students and professionals with the ability to minimise the impact of miscommunication. However, training tends to be costly and labour intensive. This pilot project aims to develop an OpenAI’s GPT-4o agent-based Chatbot with two functions: 1) to simulate cultural scenarios in medical consultations for medical students to practice their communication skills, and 2) to provide feedback on students’ performance based on the ACT Cultural Competence model, published in BEME Guide (97). The development involved one clinical communication educator, one machine learning academic, and five medical students. The medical students conducted a series of tests, and their reflections were captured to inform future development of the chatbot. Students’ evaluations underscored the value of additional learning opportunities the technology offers. Nevertheless, the authenticity of the simulator was a consistent issue. The evaluator bot provides useful feedback, which challenged participants’ cultural biases and highlighted their impact on the consultation process. Students appreciated feedback that encouraged them to acknowledge their own biases in a professional and self-aware manner. They also find the evaluator bot’s suggestions comprehensive and structured. However, it still required refinement in linguistic coherence and scenario adaptability. While further research with larger sample sizes is needed, these findings suggest that AI chatbots have the potential to serve as effective tools for training culturally competent communication in clinical settings.

Sandro Radovanović

Assistant Professor at University of Belgrade

My research interests include machine learning, development and design of decision support systems, decision theory, and fairness and justice concepts in algorithmic decision making.