Decoding medical educators’ perceptions on generative artificial intelligence in medical education

Introduction

Generative artificial intelligence (GenAI) is a transformative innovation that is rapidly integrating into various facets of society. The rapid development of generative large language models ¹ and artificial intelligence has ignited both excitement and concern in many fields, including medical education. ²

Large language models such as ChatGPT are examples of AI algorithms associated with GenAI, which, through transformer networks, are trained on massive corpuses of text to learn and reproduce language structure. These Large Language Models (LLMs) contain a cascade of transformation of the input, loosely following the principle of neural communication in the brain. ³ LLMs contain over a billion trainable parameters, which are learned and tuned from the data. ⁴

These models are useful in performing natural language processing, translation, and automatic text generation, which has led to the so-called Generative Pre-trained Transformers (GPTs). ³ Concerns about GenAI’s stem from its process of generating responses as output to user-provided prompts. It generates responses by drawing from its training across various internet resources and published texts that may contain biased or incorrect information. Devastating results may arise if the GenAI’s output is generated from erroneous sources, especially in the context of healthcare.

Previous studies have shown that medical students hold much more favorable opinions on the use of ChatGPT in medical education than graduated physicians. ⁵ However, it is unclear whether the observed differences in perception of GenAI in medical education between medical students and graduated physicians can be attributed to the particular stage of training in the medical education process or the average age difference between the two groups.

Material and methods

We developed a survey comprising 12 questions addressing faculty habits, perceived advantages, concerns, and challenges associated with GenAI in medical education. The survey was offered online to faculty members of Nova Southeastern University’s allopathic and osteopathic medical schools through Microsoft Forms. The survey was made available online to faculty in October 2023 and throughout December 2023. The data was analyzed with descriptive statistics.

Results

Forty-eight faculty members responded to the survey. Most of them had between 5 and 30 years of experience in medical education (Figure 1(a)). More than 79% of the responders expect a positive impact of GenAI on student learning (Figure 1(b)), and two-thirds (66.7%) on improving their lives (Figure 1(b)). Two-thirds of the respondents considered GenAI as important in AI changing their role as educators (Figure 1(c)). Although the majority of school faculty (85%) reported having heard of GenAI (Figure 1(d)), 42% responded to have not used it at all (Figure 1(e)).

OPEN IN VIEWER

Figure 1.

General characteristics of faculty responders. (a) Years of experience in medical education, (b) general impact, (c) impact of Generative AI (GenAI) technology on the educator role, (d) and (e) use of GenAI. x axis shows number of responders.

Generating text (33%), automating repetitive tasks (19%), and creating multimedia content (17%) were some of the common utilizations of GenAI by school faculty (Figure 1(e)).

The vast majority expressed that they prefer classes teaching to be human led rather than led by a GenAI system (Figure 2(a)). At the same time, they envisioned their class integrated with the use of Gen AI (Figure 2(a)), admitting the importance of completing tasks quicker and more efficiently (Figure 2(b)).

OPEN IN VIEWER

Figure 2.

Major benefits of Generative AI (GenAI). (a) Assessing the importance of rethinking teaching, (b) perceived advantages of using GenAI, and (c) three major curricular strengths of using GenAI. x axis shows number of responders.

The greatest perceived strengths of Gen AI were the ability to conduct more thorough background research on topics, followed by the ability to automate formative assessment and grading and make content more accessible to diverse learners (Figure 2(c)).

The faculty’s major areas of concern were regarding cheating in home assignments in assessment (97%), tendency for blunder and false information (95%), lack of context (86%), and removal of human interaction in important feedback processes (83%) (Figure 3(a)). Use of GenAI for brainstorming or reediting human written text was considered cheating by some responders to some extent, while all agreed that using the unedited output as their own would constitute cheating (Figure 3(b)). A large number of responding faculty (95.8%) believe there is an immediate need for institutional guidance and policy on the use of GenAI for students and faculty (Figure 3(c)).

OPEN IN VIEWER

Figure 3.

Concerns and challenges of Generative AI in medical education. (a) Major areas of concern, (b) cheating, and (c) the need for guardrails. x axis shows number of responders.

Discussion

The fast-paced application of GenAI holds potential in transforming healthcare education, research, and clinical practice. ⁶ This is exemplified by a study that showed the positive outlook on GenAI exhibited by medical staff at various levels within their education process. ⁵ However, a lack of patient-specific treatment plans, answers based on outdated evidence, and language barriers were some of the pitfalls associated with the use of GenAI in clinical practice. ⁵

It is clear that GenAI technology will likely change the field of medicine and healthcare. ⁷ Nonetheless, GenAI’s fast emergence brings questions among students and educators regarding learning, assessment, and integrity in scholarly endeavors. Several works have focused on proposing innovative approaches and fields of applications of GenAI in academic education; however, currently, only few studies^8–10 have focused on the perception of this technology. A recent work ¹⁰ reports the results of an international survey about the impact of GenAI on teachers and professors at all levels of education. This survey did not focus on a specific field of education. To the best of our knowledge, our survey constitutes the first effort to characterize the perception of GenAI technology among medical educators and the integration of GenAI into medical education.

Responses from our faculty reflect an optimistic and favorable outlook on GenAI’s impact on student learning. Although the study is limited by the fact that it is focused on a single institution and limited faculty response to survey, it is interesting that our survey also showed that the majority of medical educators in our institution envision that GenAI will play an important role in their educational activities. Another intriguing revelation from our survey results is that a significant portion of our survey participants was most likely of older age, with several years of experience in medical education. Despite their probable older age, they still maintain largely favorable opinions on this new technology. There was no difference in the number of years of experience in medical education between users and non-users of GenAI (data not shown).

The potential risks of bias dehumanization in the learning process are aspects that appear to be in the minds of our medical education faculty. A recent report showed that discriminatory power was statistically significantly higher in human-created questions compared to those created by ChatGPT, ¹¹ underscoring the irreplaceable aspect of the “human touch” in the didactic and assessment processes. Medical educators should foster an immersive and contextually relevant learning environment where students are invested in their learning. Academic aspects need to “humanize” learning and teaching by integrating empathy, kindness, and compassion. ¹²

The dilemma of using GPT apps in medical education poses the risk of depriving learners of the opportunity of writing a reflection on their experience when this task is delegated to the GPT. ¹³ Educators should explain to students that writing is a form of thinking and that they miss out on a critically important form of learning if they try to delegate their writing to ChatGPT or another GenAI. ¹² We need to help students learn how to use AI tools judiciously and understand their benefits and limitations.

The majority of the respondents perceived as important the capacity of GenAI to enhance learning, creativity, and efficiency, as previously highlighted in the news media. ¹⁴ A perceived advantage of GenAI in conducting more effective background research was reported by more than half of the faculty. However, it is important to note that some of these perceived strengths, such as an improved ability in conducting more efficient background research, do not hold true in reality due to GenAI not being a search engine. Students may potentially enhance their creative processes when using GenAI as a supplemental aid; however, challenges arise, and it may become detrimental in the long term when students become over-reliant on GenAI for creative content generation. Faculty can make use of these tools as means to help students with writing and research, but not as a replacement for critical thinking and original work. Faculty can also squander this opportunity when entrusting a GPT with writing a case on unprofessional behaviors to be discussed with students, as the creation of such scenario allows faculty to envision crucial issues and consider debate points. ¹³

Although GenAI models offer unparalleled benefits, such as simulating complex patient scenarios and providing personalized learning experiences, there are concerns about potential issues of accuracy, reliability of GenAI-generated content, and academic integrity.^2,15 Addressing the potential for AI-generated content to contribute to academic dishonesty appears to be a major concern among our medical educators. As GPT’s availability enables students to produce essays or assignments, both learners and educators need to realize that bypassing the learning process devaluates the educational experience.

The majority of our respondents perceived the tendency of GenAI to produce errors and biased or inappropriate information as major concerns. At the same time, GenAI content can potentially produce misinformation or biased information. ¹⁶ The fact that LLMs are stochastic machines, where the source and veracity of the data are unverifiable, is a critical issue. This aspect poses severe risks of possible misinterpretation of essential medical concepts, which would ultimately translate in a lack of trust in GenAI-generated educational material. As suggested by major guidelines in the use of LLMs, it is imperative that any information obtained from GPTs will require critical human review and the content must be verified before being used to ensure its accuracy and reliability.

Another important aspect of concern referred to the possibility that GenAI would increase surveillance or decrease privacy. The responses to our survey highlight the important role of academic institutions and governmental organizations in establishing guidelines, which are currently lacking. Students should be required to disclose whenever GenAI was used to generate content in their academic work. Equally, educators should also set an example for students and disclose their use of GenAI tools when developing educational materials.

In conclusion, as it appears that medical education faculty recognizes the potential impact of GenAI in medical education, careful deliberation of the pros and cons of GenAI is needed for its effective integration into medical education.