Co-Designing a Chatbot for Culturally Competent Clinical Communication: Experience and Reflections
Project Overview
The document explores the implementation and evaluation of an AI-driven chatbot aimed at improving culturally competent clinical communication training for medical students. It underscores the advantages of integrating generative AI into educational settings, providing scalable and cost-effective training opportunities that enable students to engage with a variety of patient profiles and receive structured feedback. The pilot study yielded promising results, showing enhancements in students' communication skills, particularly in fostering empathy and interpersonal understanding. However, it also highlighted some challenges, including the absence of non-verbal cues in the AI interactions and the repetitive nature of the feedback provided. Overall, the findings suggest that while AI technologies like chatbots can significantly contribute to medical education by simulating diverse communication scenarios, there are areas for improvement to optimize the learning experience.
Key Applications
AI-driven chatbot for clinical communication training
Context: Medical students at a UK medical school learning about culturally competent communication.
Implementation: The chatbot was co-developed using OpenAI's GPT-4o model, designed to simulate realistic patient conversations based on the ACT Cultural Competence model. It was piloted with third-year medical students who interacted with the chatbot in a low-pressure environment.
Outcomes: Students reported increased opportunities for reflection on their communication skills, improved empathy, and understanding of cultural dynamics in clinical settings. The chatbot provided structured feedback, allowing for self-paced learning.
Challenges: Limitations included the absence of non-verbal cues, the tendency of virtual patients to be overly agreeable, and the feedback sometimes being repetitive or generic.
Implementation Barriers
Technical
The chatbot's reliance on text-based interaction limits the simulation of non-verbal communication, which is crucial in real-life clinical settings.
Proposed Solutions: Future iterations could integrate descriptions of non-verbal cues or develop voice-based interactions to enhance realism.
Content Design
Feedback provided by the chatbot was sometimes perceived as repetitive and not tailored to specific scenarios, reducing its effectiveness. Implementing more dynamic behaviors and emotional variability in virtual patients could enhance engagement and realism.
Proposed Solutions: Enhancing feedback mechanisms and integrating emotional variability could lead to more personalized and effective interactions.
Project Team
Sandro Radovanović
Researcher
Shuangyu Li
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Sandro Radovanović, Shuangyu Li
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai