Potential and Design Characteristics of Artificial Intelligence-based Chatbots as Virtual Patients in Medical Education: A Review Study

Background and aims: Nowadays, there is an increasing need to learn, practice and even designmodern and technology-rich clinical environments. Meanwhile, the benefits of Virtual PatientSimulations (VPSs) for medical education have long been recognized. With recent advances inArtificial Intelligence (AI), chatbots or conversational agents have become more widespread. Aliterature review is carried out to give insight into existing educational ideas with such agents.Methods: We performed a scoping review. A literature search was performed to retrieve Englishrecords in PubMed, Scopus, and Google Scholar with three main search terms including “virtualpatient”, “chatbots” (or “conversational agents”, “dialogue agents”), and “artificial intelligence”in July ۲۰۲۲. Studies were included if they used or evaluated chatbots as virtual patients in medicaleducation. Reference lists of relevant articles were checked for further articles. Screening anddata extraction were performed in parallel by two reviewers. The included evidence was analyzedby employing the principles of thematic analysis.Results: Findings were grouped into three categories: ۱) potential applications, ۲) types of chatbotsand design characteristics, ۳) and evaluation methods. In medical education, Virtual Patients(VP) are often applied to train students in different scenarios such as recording the patient’s medicalhistory or deciding a treatment option. Conversational agents would allow a more sophisticatedand individual conversation between the student and the VP. Artificial Intelligence chatbotsallow interactive dialogue-driven teaching of medical sciences. Although chatbots offer a flexiblesolution and a vast possibility to improve healthcare education, limited examples of chatbots inHealthcare curriculum have been found.About types of chatbots, they can range from chat-like interfaces just showing an avatar of thepatient, to ۳D avatars for even more immersive and engaging interaction. Also, even though mostreceive user input through typing, some have used speech recognition. Unlike basic chatbots thatare typically rule-based, AI chatbots rely on natural language understanding, natural languageprocessing (NLP engine), and natural language generation to understand and respond to the humansthey are interacting with.Findings highlight the usefulness of simulating a VP with an AI-enabled chatbot by giving studentsthe possibility of multiple clinical cases especially, for complicated and unexpected scenariosto practice. It offers medical students an engaging and personal experience because of theinterface and the natural language. This literature review suggests that the efficacy of chatbots hasnot been thoroughly tested. Generally, the metrics used to evaluate this technology include users’usability and acceptability, accuracy, and skills evaluation.Conclusion: The exploitation of artificial intelligence, machine learning techniques, and deeplearning technologies allow for the design and development of chatbots which can be meaningfullyintegrated into medical education to develop students’ diagnostic skills.Based on the findings, conversational virtual patients may have the potential to support healthcareeducation by enabling personalized learning. Future studies could investigate the educational impact on learning outcomes and students’ satisfaction when interacting with chatbots. In addition,simultaneously with the development of chatbots in medical education, it becomes necessary todevelop national standards for designing and evaluating programs.

کلیدواژه ها:

Virtual patient ، Medical education ، Chatbot ، Artificial intelligence ، Machine learning ، Natural language processing