Applications and Challenges of Implementing Artificial Intelligence in Medical Education: Integrative Review

Abstract

Background: Since the advent of artificial intelligence (AI) in 1955, the applications of AI have increased over the years within a rapidly changing digital landscape where public expectations are on the rise, fed by social media, industry leaders, and medical practitioners. However, there has been little interest in AI in medical education until the last two decades, with only a recent increase in the number of publications and citations in the field. To our knowledge, thus far, a limited number of articles have discussed or reviewed the current use of AI in medical education.

Objective: This study aims to review the current applications of AI in medical education as well as the challenges of implementing AI in medical education.

Methods: Medline (Ovid), EBSCOhost Education Resources Information Center (ERIC) and Education Source, and Web of Science were searched with explicit inclusion and exclusion criteria. Full text of the selected articles was analyzed using the Extension of Technology Acceptance Model and the Diffusions of Innovations theory. Data were subsequently pooled together and analyzed quantitatively.

Results: A total of 37 articles were identified. Three primary uses of AI in medical education were identified: learning support (n=32), assessment of students' learning (n=4), and curriculum review (n=1). The main reasons for use of AI are its ability to provide feedback and a guided learning pathway and to decrease costs. Subgroup analysis revealed that medical undergraduates are the primary target audience for AI use. In addition, 34 articles described the challenges of AI implementation in medical education; two main reasons were identified: difficulty in assessing the effectiveness of AI in medical education and technical challenges while developing AI applications.

Conclusions: The primary use of AI in medical education was for learning support mainly due to its ability to provide individualized feedback. Little emphasis was placed on curriculum review and assessment of students' learning due to the lack of digitalization and sensitive nature of examinations, respectively. Big data manipulation also warrants the need to ensure data integrity. Methodological improvements are required to increase AI adoption by addressing the technical difficulties of creating an AI application and using novel methods to assess the effectiveness of AI. To better integrate AI into the medical profession, measures should be taken to introduce AI into the medical school curriculum for medical professionals to better understand AI algorithms and maximize its use.

Keywords: artificial intelligence; evaluation of AIED systems; medical education; real world applications of AIED systems.

Figure 1

Total publications and sum of times cited by year in the last two decades. Retrieved from Web of Science for artificial intelligence in medical education, dated April 1, 2019.

Figure 2

Search strategy for literature on the use of artificial intelligence in medical education in undergraduate, postgraduate, and specialty training in medicine and beyond (continuing medical education). ERIC: Education Resources Information Center.

Figure 3

Subgroup analysis showing the number of articles in each focus group for the target audiences.

Figure 4

Hierarchical presentation of the challenges of implementation of artificial intelligence (AI) in medical education. The upper blue rectangle shows the proportion of articles in each challenge category in the technical aspects of AI. The lower red rectangle shows the proportion of articles for challenges relating to perceived usefulness (in red) and perceived ease of use (in light red).