This study was conducted at Chiba University School of Medicine, a national university in Japan. All medical schools in Japan offer a 6-year curriculum to students entering after high school [19] and share a model core curriculum (MCC) as the nationally uniform exit competency for certifying medical degrees. The latest MCC (revised in 2022) lists the ability to recognize and use information technology, including AI, as 1 of the 10 core competencies; the achievement objectives included in the MCC define approximately two-thirds of each university’s curriculum, with the remainder allowed to be unique to each university [20].

As part of our official faculty development program, we organized a workshop for faculty and students to collaborate and invited participants from both faculty and students in August 2023. A detailed lecture on basic theory and general functions of GAI by an AI expert (EK) was conducted just prior to the workshop. Then, the participants were divided into groups of 7-8 faculty members and students to answer the above research questions. We invited faculty participants from all of the 58 departments in our school, and 47 departments agreed. Student participants were selected from students who had attended formal meetings. No participants but 1 had any experience with credited courses on AI or GAI.

We used the strength, weakness, opportunity, and threat (SWOT) framework for the workshop (Table 1) [20-23]. In analyses using the SWOT framework, the implementer identifies 4 internal or external components of stakeholders. Strengths refer to internal elements that facilitate the achievement of goals, whereas weaknesses refer to internal elements that hinder the achievement. Opportunities are external aspects that help stakeholders achieve their goals, including both positive environmental aspects and opportunities to initiate new activities. Threats are external aspects that can obstruct achieving goals [24]. The SWOT framework was first described academically by Learned et al [25] and has been used as an important tool for dealing with complex strategic situations by reducing the amount of information to improve decision-making. Specifically, it has been used to find gaps and matches between competencies and resources and the business environment [26] because it can assess alternatives and complex decision-making situations. In particular, education and health care are both major areas where SWOT has been frequently used [27]. Several examples of implementation in the medical education domain have been reported in academic journals and used for strategic planning in chaotic situations such as the COVID-19 pandemic in 2019 [28].

Following the analysis with the SWOT framework, this group work used the cross-SWOT (or TOWS [threat, opportunity, weakness, and strength]) matrix method to develop strategies (Table 2) [28,29]. Cross-SWOT analysis combines the relationships between internal and external environmental factors resulting from the SWOT analysis in a 2 × 2 grid to devise strategies for each of the 4 categories (strength and opportunity [SO], weakness and opportunity [WO], strength and threat [ST], and weakness and threat [WT]). The SO category focuses about how to exploit strengths for maximizing the potential opportunities. The ST category examines how threats can be transformed to opportunities. The WO category considers how to overcome the weaknesses with the opportunities. The WT category highlights how to avoid threats by minimizing weaknesses.

We chose the SWOT and cross-SWOT methods for this study because, first, they are suitable for our research questions since brainstorming questions can be used to reach a consensus. Further, these methods can establish strategies based on external as well as internal factors and barriers as well as facilitators. This feature is crucial for medical education, whose connection with society cannot be ignored.

All group work was tabulated and recorded on Google spreadsheets. To promote common discussions and minimize differences between groups, each group was assisted in discussions and work by 1 facilitator trained in SWOT analysis, in addition to the participants.

Qualitative content analysis was conducted to analyze the comments in the product of group work consisting of SWOT and cross-SWOT [30]. The analysis comprised the descriptions of the manifested content and interpretations of latent content [31]. Further, 4 researchers (IS, HK, KS, and NA) read the comments at the discussion and coded them to identify themes that emerged from the qualitative data independently, followed by SI checking the analysis. Points of disagreement on the data were discussed by all authors, and consensus was reached.

This study was performed following the Declaration of Helsinki and approved by the ethics committee or institutional review board at Chiba University Graduate School of Medicine (3425). All participants were informed in advance via a written document that their opinions would be recorded anonymously and analyzed collectively and that they could not withdraw after participation. The participants then communicated their consent on paper.

In this study, the impact of GAI on medical education curricula and the direction of curriculum reform based on the existence of GAI were investigated in the context of faculty development. In medical education research, the same attempt has been made by previous studies to summarize the results of SWOT analyses through qualitative analysis [21,32], which is an appropriate approach for gathering stakeholders’ opinions on the impact of an issue. Similarly, in this study, the inclusion of students in addition to medical school faculty from all areas of medical education in this study of GAI curriculum reform helped to strengthen the conclusions about GAI.

In recent years, the use of GAI in medical education has been the subject of only a few recommendations by some experts and reports of advanced practices [33,34]. However, the general state of medical education curricula has not yet been well defined, and to our knowledge, no study has yet investigated and summarized the needs of medical schools and their faculty members who manage the actual curriculum. Among the study’s findings, the need to learn about the shortcomings of GAI and some of the specific ideas for incorporating GAI into existing educational strategies were consistent with the recommendations of experts [34]. For example, Boscardin et al [33] have compiled a resource for medical educators to increase their AI literacy.

Further, the application of GAI may be particularly effective for learning in disciplines such as clinical medicine [35], where decisions are based on background knowledge backed by solid evidence. In clinical reasoning [36], for example, GAI can extend our views on a problem. Ultimately, it can add several differential diagnoses that we may not have thought of. Hence, the combined use of GAI in the clinical workplace is expected to be part of a physician’s skill set and, through training, will be expected to assist physicians in the practice [33].

Interestingly, we found a new approach that can promote the use of simulated patients in education and adaptive learning as part of “learning with GAI.” The educational use of digital patients can complement learners’ clinical experiences through experiential learning theory, providing a mechanism for information gathering and clinical decision-making in a safe environment [37]. While generally useful for understanding standardized clinical conditions [38], participants expected that the educational significance of digital patients could be amplified through the use of GAI, which can easily generate a wide variety of problems. Digital patients with GAI can even be expected to acquire some interactivity [39].

The ability to use GAI to create a variety of educational resources with less effort is also a key factor in promoting adaptive learning [40]. It is hoped that the incorporation of good practices into GAI-made educational resources will lead to the practical application of these appropriate strategies in medical education curricula. As a letter indicates, “learning with GAI” may apply to “learning about GAI” as well [41].

Furthermore, we found perspectives in this report that have not been recommended by AI experts in the past. A distinctive example is the recommendation that curricula should focus on “changes in physicians’ competencies” as the impact of GAI and “learning aside from GAI” to respond to these impacts. In areas such as medicine, where competencies of health professions are indispensable human resources, competencies that cannot be replaced by GAI, as listed from the cross-SWOT analysis in our study, will continue to be required, and as GAI becomes more prevalent in the clinical workplace, the role of physicians is expected to be focused on tasks that cannot be replaced by AI. For example, experience in the workplace cannot be replaced by GAI, nor can evaluation be faked, and hence will be emphasized more in future medical education curricula. Moreover, learners could expect to reach higher orders of cognitive domain such as “apply” and “analyze,” rather than “know” and “understand.” In this respect, evaluating GAI-generated information could also be an effective learning and assessing approach.

In terms of learners’ assessment, participants were concerned that existing assessments using students’ output may be less reliable. This suggests that summative assessments using high-stakes testing, at least in the knowledge domain, will be harder to implement. Instead, a novel assessment system has been proposed, by which the utility of assessment can complement the former approach by integrating the results of various types of feedback opportunities programmatically [42]. Such a concept of assessment may become increasingly important in future medical education curricula. Simultaneously, the importance of assessing skills and attitudes will increase because they will account for a greater proportion of a physician’s competence. These transitions are consistent with, and in fact promote, the paradigm shift in medical education, noted a decade ago [43].

Another interesting aspect of our report is that faculty and students proposed their own strategies for curriculum reform based on the impact of GAI through faculty development. Our attempt enabled faculty to engage their own intentions in the university’s curriculum reform. Classical faculty development has often adopted a top-down approach for communicating educational know-how and policies [44]. Similarly, policies for the use of GAI that have been formulated in various countries and several universities have adopted the same top-down approach. However, in curriculum reform, the usefulness of a bottom-up approach that takes advantage of faculty members’ initiative has long been pointed out [45]. Specifically, adopting a bottom-up approach facilitates faculty consensus on the curriculum and minimizes gaps in the objectives of reform and what needs to be done. In this respect, GAI is not just an educational device or technique; rather, it has the potential to revolutionize education, and its technological progress is rapid. When such an innovative technology is quickly introduced, it is essential that the entire organization is ready to embrace change [46]. If faculty development with respect to GAI is conducted by teaching expert knowledge and providing recommendations on how to (or not to) use GAI, the faculty development will have a limited effect for curriculum reform. Conversely, our faculty development program allows faculty members to formulate their own curriculum reform proposals and thus is expected to lead to the introduction of more effective ways of incorporating GAI in the context of each university. Among the items mentioned in the results of analysis, the negative impacts of GAI and the reform strategies to overcome them had much in common with public proposals, but it is significant that the faculty members themselves were able to outline their own proposals. Furthermore, Steinert et al [46] points out that faculty development is also a place of community of practice. From this perspective, it is significant that students, who are stakeholders in the learning process, are also involved, and the fact that the GAI has provided an opportunity to incorporate students’ opinions on matters that will have a large impact on their future as health professionals will give validity to the reformed curriculum. The strategies we developed were incorporated into the Guidelines on the Use of Generative AI in Teaching and Learning in our university in October 2023 (an internal document). We believe that this adoption suggests that the findings of our efforts are highly useful.

This study had several limitations. First, it was conducted at a single medical school in Japan. Since medicine is highly context dependent, future implementation of GAI in medicine may vary greatly depending on cultural and curricular characteristics. However, certain commonalities in higher education and undergraduate medical education competencies may serve as an example of how GAI may be used. Second, the number of students was smaller than that of faculty, and although the facilitator encouraged participants to pay attention to the issue of hierarchy and generate opinions during the workshop, he did not incorporate any structural devices to eliminate the issue of hierarchy between students and faculty. However, since the issue of GAI has a common impact on faculty and medical students, we do not expect that hierarchy had a significant impact on the product. Third, although the clinical workplace contains multiple health professions besides physicians, we did not incorporate the opinions of other professionals than physicians this time. In the future, similar workshops with interprofessional participants should incorporate more diverse opinions to further enhance the relevance of the developed strategies.

We conducted a qualitative analysis of the impact of GAI on medical education curricula and strategies for responding to it using the SWOT framework and cross-SWOT matrix. We recruited faculty and students to identify both positive and negative impacts of GAI on medical education curricula as well as “changes in physician specialties” as a characteristic of medical education. Curricular response principles were broadly classified into “learning about GAI,” “learning with GAI,” and “learning aside from GAI.” These principles will be the 3 pillars of medical education curriculum reform in the GAI era. Particularly, it is crucial to investigate how to maintain and promote learning aside from GAI.