Integrating Generative AI-Based Assistance Tool in Programming Education for Medical Students: A Cross-Sectional Study

Integrating Generative AI-Based Assistance Tool in Programming Education for Medical Students: A Cross-Sectional Study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Integrating Generative AI-Based Assistance Tool in Programming Education for Medical Students: A Cross-Sectional Study xiaowei xu, si zheng, lin yang, jie hao, xuwen wang, hongyu kang, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5606661/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 8 You are reading this latest preprint version Abstract Backgroud With the increasing importance of computational skills in healthcare, there is a growing need to equip medical students with programming knowledge to address complex healthcare challenges effectively. Traditional programming methods, however, can be challenging for medical students due to heavy academic loads and limited exposure to coding. Generative artificial intelligence (GenAI) presents a promising solution to these issues. Methods This research evaluated the feasibility of Chat2R, an R programming assistance tool powered by GenAI, within the postgraduate course “Practical Techniques of Medical Data Mining” at Peking Union Medical College (PUMC). A mixed-methods approach was used, combining quantitative surveys and qualitative insights to assess the tool’s effectiveness and students’ reception. Quantitative data was gathered through post-implementation surveys measuring students' perceptions of their coding proficiency and the tool’s utility. Qualitative analysis explored student interactions with Chat2R, identifying key challenges and concerns to enhance the educational experience. Results A total of 31 postgraduate students from 14 different disciplines participated in the survey. The positive feedback supported the integration of Chat2R as a valuable educational tool, highlighting GenAI’s role in enhancing computational thinking skills. Between March 13 and April 2, 2024, 28 students actively engaged with Chat2R, generating 1,603 questions. Of these, 311 were specifically related to defined data analysis processes: 138 questions on data collection, 74 on data preprocessing, 4 on data exploration, 87 on data visualization, and 8 to data modeling. Conclusion This study demonstrates that integrating Chat2R, a GenAI-based programming tool, can significantly enhance programming education for medical students. It improves computational thinking, supports both lecture and practical learning, and addresses challenges related to limited coding exposure. Positive student feedback highlights its effectiveness in providing coding assistance and fostering an interactive, student-centered learning environment. The findings also underscore the importance of professional development for educators to effectively incorporate GenAI tools into teaching. Future enhancements could include AutoML capabilities, enabling medical students to guide AI-driven data analysis and better utilize AI in clinical and research contexts. Generative AI student perception code generation programming education Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 1 INTRODUCTION Artificial intelligence (AI) is revolutionizing healthcare by enabling new approaches to analyze and interpret complex medical data. Machine learning and deep learning algorithms, in particular, have demonstrated remarkable capabilities in extracting valuable insights from multimodal data sources, including medical imaging, clinical records, genomic data, and laboratory results. These applications enhance diagnostic precision, predict patient outcomes, and support personalized treatment plans, thus facilitating real-time clinical decision-making [ 1 ]. For instance, AI-powered pathology tools can analyze tissue samples at the cellular level, aiding in the diagnosis of rare and complex conditions [ 2 , 3 ]. Similarly, in cardiology, AI models contribute to predicting patient outcomes, managing treatment plans, and guiding minimally invasive procedures using real-time data [ 4 ]. As AI becomes increasingly integral to medical practice, proficiency in AI technologies, including programming and algorithmic thinking, is crucial for medical professionals. Surveys indicate a growing demand among medical students for training in data science and AI. For example, a survey of 704 National Science Foundation investigators from the Directorate for Biological Sciences reveals that 90% were either currently analyzing or would soon be analyzing large datasets [ 5 ]. Additionally, a national survey by Stanford Medicine indicated that medical students’ interest in supplemental education in coding and AI-related topics [ 6 ]. However, acquiring such skills remains challenging due to intensive curricular demands and limited exposure to computational concepts. Medical students face several specific challenges in AI-related programming education. These include difficulty understanding abstract programming logic without sufficient foundational training, limited opportunities for hands-on practice, and lack of real-time guidelines during coding exercises. Moreover, conventional instructional methods, often based on static lectures or textual assignments [ 7 ], do not cater to individual learning needs and can lead to frustration or disengagement. Generative Artificial Intelligence (GenAI), particularly models based on large language models (LLMs), offers a promising solution to these educational challenges. Tools such as GitHub Copilot [ 8 ] and ChatGPT [ 9 ] enable natural language-based code generation and explanation, thereby lowering the learning curve and offering on-demand support. These technologies have been shown to alleviate programming anxiety, promote inquiry-based learning, and support students with diverse computational backgrounds by offering immediate, personalized feedback, and interactive problem-solving capabilities [ 10 ]. Despite these advancements, limited research has examined how medical students perceive and interact with GenAI tools in the context of programming education. Previous studies have primarily compared GenAI capabilities with those of human experts, with little attention paid to learner-centered evaluation [ 11 – 13 ]. To address this gap, we developed Chat2R, a GenAI-powered programming assistant specifically designed to support medical students in acquiring programming and AI-related skills. Chat2R improves upon traditional instruction by offering medical students real-time, natural language-driven support in three critical areas: code generation, code explanation, and code debugging. Unlike conventional teaching methods, which often rely on static lecture content and delayed feedback, Chat2R enables students to interactively generate R code from plain language prompts, receive instant explanations to enhance their understanding, and iteratively debug code with AI assistance. These features make programming more accessible, adaptive, and student-centered. Furthermore, Chat2R accommodates varied learning paces, allowing students to progress according to their comfort level, thereby reducing anxiety and promoting confidence. We conducted a cross-sectional study using a mixed-methods approach to evaluate Chat2R’s effectiveness, usability, and impact on students’ learning experience. 2 METHOD 2.1 Participants This cross-sectional study was conducted with postgraduate students enrolled in the elective course "Practical Techniques of Medical Data Mining" (Course No. INSC11011) at Peking Union Medical College (PUMC) [ 14 ]. At the beginning of the semester, 50 students from various disciplines registered for the course. All enrolled students were invited to participate in the study, and participation was entirely voluntary. No exclusion criteria were applied. Ultimately, 31 students who attended the course and completed the questionnaire were included in the analysis. The other students either chose not to participate or did not use the GenAI-based programming assistance tool, Chat2R, and were therefore excluded from the final dataset. 2.2 Design of the programming assistance tool: Chat2R Chat2R is a programming assistance tool that leverages generative artificial intelligence techniques to streamline the coding process for medical students. The tool assists with three primary functions: code generation, code explanation, and code debugging. It automates the creation of code based on user’s natural language inputs, reducing the time and effort required for students to write code from scratch. Additionally, it provides detailed explanations of generated code to help students understand the logic and structure behind the code, and it assists students in identifying and correcting errors in their code, thereby improving their problem-solving skills and understanding of programming concepts. The flowchart in Fig. 2 illustrates the implementation and workflow of Chat2R, the code generation tool leveraging the Qwen-72B-Chat [ 15 ] interface to assist in learning R programming. Chat2R starts by receiving a problem description and creating an initial prompt to generate R code. Configured through a system prompt, Chat2R produces the code, which the user then executes. If an error occurs during execution, Chat2R generates a new prompt to revise and improve the code. This iterative process of generating prompts, refining code, and user execution continues until the code achieves the expected results, effectively solving the given problem. The design of Chat2R focuses on user-friendliness and educational value, ensuring seamless integration into the medical data mining course. The tool is accessible via a web interface, allowing students to interact with it using natural language queries. 2.3 Design of the questionnaire A 26-item questionnaire was developed based on relevant literature and refined through consultation with two medical artificial intelligent education experts to ensure content validity [ 16 ][ 17 ]. It comprised four sections: (1) demographic information (Q1-Q4); (2) knowledge and familiarity with GenAI (Q5-Q11); (3) perceptions and experiences using Chat2R (Q12-Q18); (4) perspectives on GenAI’s potential role in clinical practice (Q19-Q24). The questionnaire included multiple item types. Items Q5-Q11 and Q15-Q24 were measured using a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree), designed to assess students’ familiarity with GenAI technologies and their attitudes toward their application in education and clinical practice. Other items (e.g., Q12-Q14) utilized frequency-based or binary response formats (e.g., Yes/No) to capture students’ usage behaviors and overall impressions of Chat2R. The questionnaire also included two open-ended questions to capture quantitative feedback. A pilot test with five postgraduate students was conducted to assess clarity, comprehensibility, and item coherence. Based on the feedback, minor revisions were made to improve wording and structure. To evaluate internal consistency, Cronbach’s alpha was calculated using the full dataset from the main study (n = 31). The resulting alpha coefficient was 0.73, indicating acceptable reliability of the instrument. 2.4 Data collection process The data collection process involved multiple stages to ensure comprehensive and reliable data. The questionnaire was administered to students’ post-implementation of Chat2R, ensuring that they had sufficient time to interact with the tool. Instructors observed student interactions with Chat2R during the course, noting any challenges or areas where students required additional support. Detailed logs of student interactions with Chat2R were collected, capturing the types of queries and issues encountered. 2.5 Analysis The analysis of the collected data was conducted through a multi-stage approach to ensure a comprehensive evaluation of Chat2R’s effectiveness in medical education. Initially, descriptive statistics were applied to the survey responses to summarize students’ perceptions regarding Chat2R's ease of use, educational value, and overall usefulness. To gain deeper insights into the qualitative data obtained from open-ended responses, thematic analysis was conducted, involving the systematic coding of responses to identify recurring themes and patterns related to students' experiences and challenges when using Chat2R. The quantitative and qualitative findings were then integrated to provide a holistic assessment of Chat2R’s impact, using cross-validation to compare survey results with qualitative insights and ensure consistency and reliability. Additionally, logs of student interactions with Chat2R were analyzed to classify the types of queries submitted (e.g., code generation, code explanation, code debugging) and to identify common issues faced by students. This mixed-methods approach allowed for a nuanced understanding of how Chat2R supports the learning process and highlighted areas for further refinement and development. 2.6 Ethics The Ethics Committee of the Institute of Medical Information, Chinese Academy of Medical Sciences and PUMC, granted approval for this study (IMICAMS/01/20/HREC). All participants were informed that their responses would contribute to public-facing research, and written informed consent was obtained prior to participation. All procedures were conducted in strict adherence to the principles outlined in the Declaration of Helsinki. 3 RESULTS 3.1 Participant Demographics A total of 31 postgraduate students completed the questionnaire. Participants represented 14 academic disciplines, with 18 females (58.06%) and 13 males (41.94%). The average age was 25 years (range: 20–44). This diverse sample included students from fields such as medical informatics, internal medicine, public health, and nursing. Table 1 provides a detailed breakdown of these demographics, illustrating the multidisciplinary nature of the participants involved in the research. Table 1 Participants’ demographic information Characteristic N(%) % Sex Male 13 41.94 Female 18 58.06 Major Medical Informatics 6 19.4 Internal Medicine 4 12.9 Biomedical Engineering 4 12.9 Information Science 3 9.7 Epidemiology and Health Statistics 3 9.7 Clinical Diagnostics 3 9.7 Imaging and Nuclear Medicine 1 3.2 Geriatrics 1 3.2 Public Health 1 3.2 Nursing 1 3.2 Pediatrics 1 3.2 Clinical Anesthesiology 1 3.2 Clinical Integration of Traditional Chinese and Western Medicine 1 3.2 Clinical Medicine 1 3.2 3.2 Knowledge and familiarity with GenAI Technologies Students reported a generally high level of awareness and understanding of GenAI technologies. The average composite score for Q5-Q11 was 4.32 (SD = 0.48) on a 5-point Likert scale (1 = strongly disagree, 5 = strongly agree). The highest-rated item was Q7: “I understand generative AI technologies like ChatGPT can generate output that is factually inaccurate” (Mean = 4.35, SD = 0.55), while the lowest-rated item was Q9: “I understand generative AI technologies like ChatGPT can exhibit biases and unfairness in their output" (Mean = 3.65, SD = 0.88). Students from technical disciplines such as epidemiology, medical informatics, and biomedical engineering scored higher on GenAI knowledge than those from clinical majors (see Fig. 3 ). 3.3 Chat2R Usage and Perceived Usefulness Among the 31 students surveyed, 23 (74.2%) reported prior experience with GenAI products before the course (Q13), and an equal number (23, 74.2%) used Chat2R during the course. Chi-square analysis revealed no significant association between prior GenAI experience and actual usage of Chat2R during class sessions (χ² (1, N = 31) = 0.0, p > 0.05), suggesting that prior exposure to GenAI tools was not a determining factor for Chat2R engagement. In terms of usage frequency (Q14), over half of Chat2R users (13, 56.52%) reported using the tool weekly, while the remaining 43.48% used it occasionally. Most users found the tool easy to use (Q15): 8 students (34.78%) rated it as “very easy”, 14 students (60.87%) as “easy”, and 1 (4.35%) as “medium”. However, no significant association was found between usage frequency and perceived ease of use (χ² (2, N = 23) = 2.65, p > 0.05). When asked whether Chat2R helped their learning (Q16), a large majority responded positively: 12 students (38.71%) rated it as “very helpful,” and 18 (58.06%) as “helpful,” with only one student selecting “medium.” Subgroup analysis showed that all 23 Chat2R users considered the tool either “very helpful” or “helpful.” Among non-users, most still perceived the tool as helpful based on peer observation or classroom context, though their ratings were slightly less enthusiastic. Students’ overall impressions of Chat2R (Q17) were also favorable. 13 students (41.94%) rated the tool as “very good,” 16 (51.61%) as “good,” and 2 (6.45%) as “medium.” Interestingly, students who used Chat2R tended to have a more uniformly positive impression, while responses from non-users were more varied. A statistically significant difference in overall impressions between users and non-users was observed (χ² test, p < 0.05). Regarding future use intentions (Q18), most students expressed willingness to continue using Chat2R. Specifically, 15 (48.39%) were “very willing,” 12 (38.71%) “willing,” 3 (9.68%) “neutral,” and only one (3.23%) was “not willing.” That single student who expressed unwillingness had not used Chat2R in class. However, there was no statistically significant association between prior usage and future willingness to use the tool (χ² (3, N = 31) = 3.92, p > 0.05). Overall, these results suggest that Chat2R was perceived as a useful, accessible, and well-received learning tool by most students, including those with no prior coding or AI experience. 3.4 Perspectives on GenAI’s Potential Role in Clinical Practice Students generally expressed strong support for the integration of GenAI into future clinical workflows. On the 5-point Likert scale, the highest-rated item was Q21: “ Generative AI technologies can help me save time ” (Mean = 4.65, SD = 0.49), indicating a strong belief in the efficiency-enhancing potential of these tools. The lowest-rated item was Q20: “ Generative AI technologies can improve my computational thinking skills ” (Mean = 3.65, SD = 0.66), suggesting some uncertainty about the long-term impact of GenAI on students’ technical skill development. Students also expressed high agreement with statements reflecting the practical utility of GenAI tools: the ability to offer personalized and immediate feedback (Mean = 4.23, SD = 0.56), generate unique insights (Mean = 4.32, SD = 0.65), and provide around-the-clock availability (Mean = 4.35, SD = 0.75). These findings indicate that students perceive GenAI as a valuable supplementary assistant in clinical practice. Further analysis of willingness scores across different academic disciplines showed notable variation. Students majoring in Medical Informatics (Median = 27) and Epidemiology and Health Statistics (Median = 26) demonstrated the highest levels of willingness to adopt GenAI technologies in future practice. In contrast, students from Clinical Diagnostics reported the lowest median score (Median = 22), as illustrated in Fig. 4 . 3.5 Qualitative Findings Open-ended responses provided additional insights into students’ perceptions of Chat2R. When asked about its role in supporting their learning, many students highlighted Chat2R’s utility in assisting with code correction and debugging. Several respondents noted that the tool helped them identify and resolve errors, clarify coding logic, and simplify complex expressions. In their words, Chat2R “ corrected errors, explained code functions, and simplified logic ” and “ provided coding ideas and alternative approaches ”. Beyond technical support, students also described Chat2R as facilitating clearer thinking and offering diverse perspectives when tackling programming tasks. For instance, one student remarked that the tool “ helped structure thoughts and explore different problem-solving strategies ”. In addition, Chat2R was perceived to enhance learning efficiency and promote independent exploration, with multiple students commenting on its ability to “ improve speed and autonomy ” when writing and modifying code. Students also offered suggestions for improving the tool and its implementation. Several participants expressed a desire for extended access to Chat2R after the course concluded, while others recommended incorporating more practical case-based examples and enabling file or image uploads to support real-world data analysis scenarios. Additional suggestions included improving the user interface, such as easier access to chat history, and expanding compatibility with other statistical tools. A few students also proposed exploring advanced AI topics, including deep learning, in future iterations of the tool. Collectively, these qualitative insights underscore Chat2R’s perceived effectiveness as a supportive educational tool while also highlighting areas for refinement to better align with learners’ evolving needs. 3.6 Interaction Behavior Analysis in Chat2R To better understand how students engaged with the Chat2R tool, interaction logs were analyzed with a focus on session structure, query patterns, content relevance, and input characteristics. As shown in Fig. 5 , most chat sessions consisted of a single round of interaction, indicating that most students used Chat2R to obtain quick answers rather than to engage in extended, multi-turn conversation. This suggested that the tool was primarily used for task-specific assistance. To further characterize usage patterns, the proportion of follow-up versus new questions was examined. As illustrated in Fig. 6 , follow-up questions occurred more frequently than entirely new prompts. This suggests an iterative learning behavior in which students used Chat2R to refine previous outputs or clarify ongoing tasks, rather than starting with unrelated queries each time. Next, the questions were categorized by their relevance to different stages of the data analysis process, including data collection, preprocessing, exploration, visualization, and modeling. Of the 1,603 questions submitted, 311 could be directly mapped to a defined analysis phase: 138 related to data collection, 74 to preprocessing, 4 to exploration, 87 to visualization, and 8 to modeling (Fig. 7 ). The remaining questions focused primarily on general code explanation, debugging, and syntax clarification, and thus did not correspond to a single analytical stage. Finally, we examined the distribution of question lengths to better understand students’ prompting behavior and potential reliance on the tool. As shown in Fig. 8 , most queries were relatively brief-1,457 of 1,603 questions contained fewer than 200 characters. The most frequent input length was 11 characters (n = 74), with question lengths ranging from 1 to 1,375 characters. These findings indicate that students tend to use short, targeted queries to interact with Chat2R, consistent with task-oriented and efficiency-driven usage. 4 DISCUSSIONS This research systematically evaluated the usability and perceived effectiveness of large language models (LLMs), specifically Chat2R, as an assistance tool for programming education among medical students. A mixed-methods approach was adopted, combining quantitative survey results with qualitative insights to assess both user experience and the pedagogical value of GenAI technologies. The findings demonstrated that Chat2R provided practical benefits in the context of medical programming education. Students reported that the tool facilitated learning through real-time code generation, explanation, and debugging, which aligns with previous research suggesting that AI-powered educational technologies can increase learner engagement, lower cognitive burden, and promote programming confidence [ 17 – 19 ]. These results are consistent with studies indicating that immediate feedback and AI-enabled guidelines improve student autonomy and reduce frustration in learning complex skills [ 20 ]. By incorporating Chat2R into courses that focus on data analysis and medical programming, educational institutions can facilitate interactive, student-centered learning that helps students overcome common obstacles in programming. This integration could mitigate the frustration often associated with traditional programming instruction [ 21 ]. Additionally, implementing Chat2R in flipped classroom models, where students use the tool for coding practice at home, would allow in-class time to be devoted to deeper learning through collaborative problem-solving and discussion. Chat2R’s real-time coding assistance can be extended beyond the classroom to include broader student support systems, such as coding labs and tutoring services. This broader implementation would offer students additional resources for independent learning and promote the development of problem-solving skills essential for medical practice. Studies on supplemental learning resources have shown that access to support tools outside of classroom hours contributes to a deeper understanding of course material and improves academic performance [ 22 ]. Deploying Chat2R in supplementary learning contexts could provide personalized feedback aligned with students’ varying levels of proficiency. Research on adaptive learning technologies highlights that tailored support enhances learning retention and student satisfaction [ 20 ]. The tool's capability to cater to different learning paces means that students requiring more time can receive targeted help, while advanced learners can explore complex concepts at their own speed. This flexibility addresses the diversity of programming experience among medical students, supporting them in building foundational and advanced skills that are increasingly essential in medical research and practice. Despite its advantages, concerns remain regarding the potential for over-reliance on Chat2R. This aligns with earlier studies cautioning that excessive dependence on AI-based educational tools may foster surface-level learning and hinder the development of independent problem-solving skills [ 23 ][ 24 ]. To mitigate such risks, it is essential that GenAI tools be integrated within structured and pedagogically sound learning environments that emphasize guided instruction, critical reflection, and active engagement with the learning tasks. Effective implementation of Chat2R and similar tools also depends on targeted faculty development. Instructors require training not only in the technical use of GenAI platforms but also in evidence-based strategies for facilitating productive student-AI interaction [ 25 ]. Prior research has demonstrated that professional development in educational technology enhances instructors’ capacity to design adaptive, inclusive, and student-centered learning environments [ 26 , 27 ]. Institutional support is equally important to ensure that GenAI integration aligns with curricular goals, adheres to ethical standards, and maintains technical reliability. This study has several limitations that must be acknowledged. Most notably, the absence of a control group limits the ability to draw causal inferences. Although some students did not actively use Chat2R, the study was not designed to compare their outcomes systematically with those of users. Furthermore, reliance on self-reported data introduces potential biases, including recall inaccuracies and social desirability effects. Future research should address these limitations by employing randomized controlled trials or longitudinal study designs, enabling the tracking of learning outcomes and perceptions over time. Incorporating objective performance indicators, such as coding assessments or log-based learning analytics, would provide a more comprehensive evaluation of Chat2R’s educational impact. Looking ahead, the functionality of Chat2R could be further expanded through the integration of AutoML [ 28 ](automated machine learning) capabilities. This enhancement would allow students to transition from manual coding tasks to higher-order activities such as model interpretation and clinical insights generation. Such development aligns with broader trends in which AI augments, rather than replaces human expertise in health data analysis [ 29 ]. By enabling learners to move from implementation to strategic oversight, Chat2R has the potential to prepare future medical professionals to critically supervise AI-driver tools in both research and clinical contexts. In summary, Chat2R demonstrates strong potential as a generative AI–powered educational tool for supporting programming education among medical students. With thoughtful instructional design, professional development, and institutional support, such tools can become integral components of modern medical education, fostering more personalized, interactive, and effective learning environments. 5 CONCLUSIONS This study demonstrates that integrating Chat2R, a generative AI-based code generation tool, into the "Practical Techniques of Medical Data Mining" course effectively supported medical students in acquiring programming skills. Students reported a positive learning experience, noting enhanced computational thinking and reduced barriers to engaging with coding tasks. Analysis of usage patterns revealed diverse learning needs, with a strong emphasis on code generation and moderate reliance on debugging support. These findings suggest that GenAI tools such as Chat2R have the potential to improve the accessibility and personalization of programming education in medical training. Further research should explore the long-term effects of such technologies on learning outcomes across diverse educational settings. Declarations Conflict of Interest Statement The authors declare that they have no conflict of interest. Author Contribution Xiaowei Xu: Methodology, Software, Data curation, Investigation, Writing - original draft. Zheng Si: Validation, Investigation. Lin Yang: Validation. Jie Hao: Methodology. Xuwen Wang: Data curation. Hongyu Kang: Validation. Chao Ma: Supervision. Xudong Lv: Supervision. Jiao Li: Funding acquisition, Conceptualization, Supervision. All authors contributed to the development of this manuscript and approved the final version for submission. Acknowledgement We gratefully appreciate all of the participants and staff for their contributions. Data Availability The datasets generated during the current study are not publicly available due to participant confidentiality and institutional restrictions, but are available from the corresponding author on reasonable request. References Ahsan MM, Luna SA, Siddique Z. Machine-learning-based disease diagnosis: a comprehensive review. Healthcare. 2022;10(3):541. 10.3390/healthcare10030541 . Shafi S, Parwani AV. Artificial intelligence in diagnostic pathology. Diagn Pathol. 2023;18(1):109. 10.1186/s13000-023-01375-z . Huang Z, Yang E, Shen J et al. A pathologist-AI collaboration framework for enhancing diagnostic accuracies and efficiencies. Nat Biomed Eng. 10.1038/s41551-024-01223-5 Gala D, Behl H, Shah M, Makaryus AN. The Role of Artificial Intelligence in Improving Patient Outcomes and Future of Healthcare Delivery in Cardiology: A Narrative Review of the Literature. Healthc (Basel). 2024;12(4):481. 10.3390/healthcare12040481 . Barone L, Williams J, Micklos D. Unmet needs for analyzing biological big data: A survey of 704 NSF principal investigators. PLoS Comput Biol. 2017;13(10):e1005755. 10.1371/journal.pcbi.1005755 . Stanford Medicine 2020 Health Trends Report. The Rise of the Data-Driven Physician. Stanford, CA: Stanford Medicine URL: https://med.stanford.edu/content/dam/sm/school/documents/Health-Trends-Report /Stanford Medicine Health Trends Report 2020.pdf [accessed 2024-09-16]. Yang L, Zheng S, Xu X, et al. Medical Data Mining Course Development in Postgraduate Medical Education: Web-Based Survey and Case Study. JMIR Med Educ. 2021;7(4):e24027. 10.2196/24027 . GitHub Copilot. URL: https://github.com/features/copilot [accessed 2024-11-16]. ChatGPT. URL:https://chat.openai.com/ [accessed 2024-11-16]. Tayebi Arasteh S, Han T, Lotfinia M, et al. Large language models streamline automated machine learning for clinical studies. Nat Commun. 2024;15(1):1603. 10.1038/s41467-024-45879-8 . Fan A, Gokkaya B, Harman M et al. Large language models for software engineering: Survey and open problems. In: 2023 IEEE/ACM International Conference on Software Engineering: Future of Software Engineering (ICSE-FoSE); 2023:31–53. 10.1109/ICSE-FoSE59343.2023.00008 Chang Y, Wang X, Wang J et al. A survey on evaluation of large language models. ACM Trans Intell Syst Technol. 2024;15(3): Article 39. 10.1145/3641289 Kang K, Yang Y, Wu Y, Luo R. Integrating Large Language Models in Bioinformatics Education for Medical Students: Opportunities and Challenges. Ann Biomed Eng. 2024;52(9):2311–5. 10.1007/s10439-024-03554-5 . Chinese Academy of Medical Sciences and Peking Union Medical College. URL: http://english.cams.cn/index.html [accessed 2024-11-16]. Bai J, Bai S, Chu Y et al. Qwen technical report. arXiv. Preprint posted online September 28, 2023. Arxiv: 2309.16609. Mai D, Da C, Hanh N. The use of ChatGPT in teaching and learning: a systematic review through SWOT analysis approach. Front Educ. 2024;9:1328769. 10.3389/feduc.2024.1328769 . Chan C, Hu W. Students’ voices on generative AI: perceptions, benefits, and challenges in higher education. Int J Educ Technol High Educ. 2023;20:43. 10.1186/s41239-023-00411-8 . Hooda M, Rana C, Dahiya O et al. Artificial Intelligence for Assessment and Feedback to Enhance Student Success in Higher Education. Math Probl Eng. 2022; 2022: 1–19. 10.1155/2022/5215722 Labadze L, Grigolia M, Machaidze L. Role of AI chatbots in education: systematic literature review. Int J Educ Technol High Educ. 2023;20:56. 10.1186/s41239-023-00426-1 . Taylor D, Yeung M, Bashet A. Personalized and Adaptive Learning. In: Ryoo J, Winkelmann K, editors. Innovative Learning Environments in STEM Higher Education. Springer Briefs in Statistics. Cham: Springer; 2021. 10.1007/978-3-030-58948-6_2 . Weng X, Chiu T. Instructional design and learning outcomes of intelligent computer assisted language learning: Systematic review in the field. Comput Educ Artif Intell. 2023;4:100117. 10.1016/j.caeai.2022.100117 . Li Z, Chen H, Gao X. The influencing factors and predictability of primary school students’ learning performance in online supplementary classes. Educ Inf Technol. 2024;29:10995–1021. 10.1007/s10639-023-12246-x . Gerlich M. AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies. 2025;15(1):6. https://doi.org/10.3390/soc15010006 . Lee H, Lee HP, Sarkar A, Tankelevitch L et al. The impact of generative AI on critical thinking: self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. Proc ACM CHI Conf Hum Factors Comput Syst. 2025. Mulyani H, Istiaq M, Shauki E, et al. Transforming education: exploring the influence of generative AI on teaching performance. Cogent Educ. 2025;12(1). 10.1080/2331186X.2024.2448066 . Uzorka A, Namara S, Olaniyan A. Modern technology adoption and professional development of lecturers. Educ Inf Technol. 2023;28:14693–719. 10.1007/s10639-023-11790-w . Zhang L, Basham J, Carter R, et al. Exploring Factors associated with the implementation of student-centered instructional practices in U.S. classroom. Teach Teach Educ. 2021;99:103273. 10.1016/j.tate.2020.103273 . AutoML.org. AutoML: Automated Machine Learning. https://www.automl.org/automl/ . [accessed 2024-11-16]. Yu K, Healey E, Leong T, et al. Medical Artificial Intelligence and Human Values. N Engl J Med. 2024;390(20):1895–904. 10.1056/NEJMra2214183 . Additional Declarations No competing interests reported. Supplementary Files SurveyGenerativeAIMedEd.docx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 21 Apr, 2025 Reviews received at journal 21 Apr, 2025 Reviews received at journal 13 Apr, 2025 Reviewers agreed at journal 12 Apr, 2025 Reviewers agreed at journal 10 Apr, 2025 Reviewers invited by journal 09 Apr, 2025 Submission checks completed at journal 07 Apr, 2025 First submitted to journal 07 Apr, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5606661","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":440965708,"identity":"841c439b-e2cd-4238-a093-8f8096814eb1","order_by":0,"name":"xiaowei xu","email":"","orcid":"","institution":"College of Biomedical Engineering \u0026 Instrument Science, Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"xiaowei","middleName":"","lastName":"xu","suffix":""},{"id":440965709,"identity":"072b6db5-155f-4b69-b82b-0616a40c53fd","order_by":1,"name":"si zheng","email":"","orcid":"","institution":"Institute of Medical Information \u0026 Library, Chinese Academy of Medical Sciences \u0026 Peking Union Medical College","correspondingAuthor":false,"prefix":"","firstName":"si","middleName":"","lastName":"zheng","suffix":""},{"id":440965710,"identity":"9a0c5e8c-0d1c-4823-91d2-1ee75e4968e7","order_by":2,"name":"lin yang","email":"","orcid":"","institution":"Institute of Medical Information \u0026 Library, Chinese Academy of Medical Sciences \u0026 Peking Union Medical College","correspondingAuthor":false,"prefix":"","firstName":"lin","middleName":"","lastName":"yang","suffix":""},{"id":440965711,"identity":"6732a585-2450-4813-b91a-6ff379860073","order_by":3,"name":"jie hao","email":"","orcid":"","institution":"Institute of Medical Information \u0026 Library, Chinese Academy of Medical Sciences \u0026 Peking Union Medical College","correspondingAuthor":false,"prefix":"","firstName":"jie","middleName":"","lastName":"hao","suffix":""},{"id":440965712,"identity":"bc8caece-93a2-4504-8027-ca7778dc571e","order_by":4,"name":"xuwen wang","email":"","orcid":"","institution":"Institute of Medical Information \u0026 Library, Chinese Academy of Medical Sciences \u0026 Peking Union Medical College","correspondingAuthor":false,"prefix":"","firstName":"xuwen","middleName":"","lastName":"wang","suffix":""},{"id":440965713,"identity":"94ad7f2a-e2a8-4ce5-95d4-79df697fef65","order_by":5,"name":"hongyu kang","email":"","orcid":"","institution":"Institute of Medical Information \u0026 Library, Chinese Academy of Medical Sciences \u0026 Peking Union Medical College","correspondingAuthor":false,"prefix":"","firstName":"hongyu","middleName":"","lastName":"kang","suffix":""},{"id":440965714,"identity":"1bfb5bbe-07f7-41db-ac27-a8f43d108dea","order_by":6,"name":"chao ma","email":"","orcid":"","institution":"Institute of Basic Medical Sciences, Department of Human Anatomy, Histology and Embryology, Chinese Academy of Medical Sciences, School of Basic Medicine, Peking Union Medical College","correspondingAuthor":false,"prefix":"","firstName":"chao","middleName":"","lastName":"ma","suffix":""},{"id":440965715,"identity":"94302124-6a5c-4032-ae52-658b598eaa42","order_by":7,"name":"xudong lu","email":"","orcid":"","institution":"College of Biomedical Engineering \u0026 Instrument Science, Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"xudong","middleName":"","lastName":"lu","suffix":""},{"id":440965716,"identity":"7d445694-c310-4ad0-9dd7-7a72bed110a9","order_by":8,"name":"jiao li","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0klEQVRIiWNgGAWjYBACPmYGBoMEBgY5NgbGBuK0sEG1GJOgBUonEqkepIWdgaHgQY1Nep/Y4cYPP3PqGPjbDzB+LiDosGNpuW3Sic2SvdsOM0icSWCWnkFQC9thkJY2Bt5tBxgYbgAFeQhq+Xc4nQ2ohfHvtjoGeaK0JLYdTgBpYebdxsxgQFgLY4NBYl+aIcgv0rLbDvMYngEy8Gnh5z98zPDHNxt5+dnpDz++3VYnJ3f88MHP+LQwMDC2GSBzgYoJxynzA0IqRsEoGAWjYIQDANxHPa2sz1s1AAAAAElFTkSuQmCC","orcid":"","institution":"Institute of Medical Information \u0026 Library, Chinese Academy of Medical Sciences \u0026 Peking Union Medical College","correspondingAuthor":true,"prefix":"","firstName":"jiao","middleName":"","lastName":"li","suffix":""}],"badges":[],"createdAt":"2024-12-09 07:23:25","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5606661/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5606661/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":80578852,"identity":"a834977c-0920-436e-ad2c-743f892a6e43","added_by":"auto","created_at":"2025-04-14 23:05:57","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":64580,"visible":true,"origin":"","legend":"\u003cp\u003eDesign of the study\u003c/p\u003e","description":"","filename":"floatimage126.png","url":"https://assets-eu.researchsquare.com/files/rs-5606661/v1/a4e38f141eeb9ff185f13079.png"},{"id":80578853,"identity":"7d89a04d-1691-4ab2-9036-7fca63f5a89c","added_by":"auto","created_at":"2025-04-14 23:05:57","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":74532,"visible":true,"origin":"","legend":"\u003cp\u003eImplementation and workflow of Chat2R\u003c/p\u003e","description":"","filename":"floatimage213.png","url":"https://assets-eu.researchsquare.com/files/rs-5606661/v1/895e956ae146ffb36d271d9f.png"},{"id":80577596,"identity":"6f21c1da-e8d4-4f96-a10f-b326272e150d","added_by":"auto","created_at":"2025-04-14 22:57:57","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":55181,"visible":true,"origin":"","legend":"\u003cp\u003eMedian knowledge scores on generative AI technologies by academic discipline. Students from technical disciplines (e.g., epidemiology, medical informatics, biomedical engineering) demonstrated higher self-reported knowledge scores compared to those from clinical majors.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-5606661/v1/b88f3e93b2cd0d0d0784a6ff.png"},{"id":80581601,"identity":"10f75628-4964-42b9-85f1-cf8a70a1dce1","added_by":"auto","created_at":"2025-04-14 23:21:57","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":53974,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of students’ willingness scores toward GenAI adoption in clinical practice, grouped by academic discipline. The highest median scores were observed among students in Medical Informatics and Epidemiology and Health Statistics, whereas Clinical Diagnostics students showed lower overall willingness. This may reflect discipline-specific familiarity with digital tools or perceived relevance of GenAI applications.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-5606661/v1/ded8230abea017b59f80bb42.png"},{"id":80580198,"identity":"e72cf933-7be7-42ba-bfc9-f5c67b3c7c7a","added_by":"auto","created_at":"2025-04-14 23:13:57","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":22311,"visible":true,"origin":"","legend":"\u003cp\u003eNumber of interactions rounds per chat session.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-5606661/v1/0f2d951034adbc9c31a78111.png"},{"id":80578856,"identity":"d572f3a9-071f-4381-af44-693d60a907e9","added_by":"auto","created_at":"2025-04-14 23:05:57","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":31894,"visible":true,"origin":"","legend":"\u003cp\u003eFrequency of follow-up questions versus new questions per chat session.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-5606661/v1/f67fbed45ea44e0f1745b81d.png"},{"id":80578861,"identity":"08c01773-61de-4f5a-954c-30521287359c","added_by":"auto","created_at":"2025-04-14 23:05:57","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":40619,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of Chat2R questions by stage of the data analysis process.\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-5606661/v1/bc0b84c8237d7627ceb1a652.png"},{"id":80577605,"identity":"c0bc05e5-cab5-4c86-a918-ddfa1a0d8154","added_by":"auto","created_at":"2025-04-14 22:57:57","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":20879,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of question length (in characters) submitted to Chat2R.\u003c/p\u003e","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-5606661/v1/2cf2c6b76133c09761762c72.png"},{"id":80582409,"identity":"8fbdbce4-e897-49ab-9de7-bd0f567d49ef","added_by":"auto","created_at":"2025-04-14 23:29:57","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1038443,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5606661/v1/8177889a-7f80-4ed3-9b98-24867e970c8d.pdf"},{"id":80581602,"identity":"0f6afdfd-cab0-4efd-aef5-eab15d2f397a","added_by":"auto","created_at":"2025-04-14 23:21:57","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":28493,"visible":true,"origin":"","legend":"","description":"","filename":"SurveyGenerativeAIMedEd.docx","url":"https://assets-eu.researchsquare.com/files/rs-5606661/v1/55bc85c212f0ed12566cc494.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Integrating Generative AI-Based Assistance Tool in Programming Education for Medical Students: A Cross-Sectional Study","fulltext":[{"header":"1 INTRODUCTION","content":"\u003cp\u003eArtificial intelligence (AI) is revolutionizing healthcare by enabling new approaches to analyze and interpret complex medical data. Machine learning and deep learning algorithms, in particular, have demonstrated remarkable capabilities in extracting valuable insights from multimodal data sources, including medical imaging, clinical records, genomic data, and laboratory results. These applications enhance diagnostic precision, predict patient outcomes, and support personalized treatment plans, thus facilitating real-time clinical decision-making [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. For instance, AI-powered pathology tools can analyze tissue samples at the cellular level, aiding in the diagnosis of rare and complex conditions [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Similarly, in cardiology, AI models contribute to predicting patient outcomes, managing treatment plans, and guiding minimally invasive procedures using real-time data [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAs AI becomes increasingly integral to medical practice, proficiency in AI technologies, including programming and algorithmic thinking, is crucial for medical professionals. Surveys indicate a growing demand among medical students for training in data science and AI. For example, a survey of 704 National Science Foundation investigators from the Directorate for Biological Sciences reveals that 90% were either currently analyzing or would soon be analyzing large datasets [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Additionally, a national survey by Stanford Medicine indicated that medical students\u0026rsquo; interest in supplemental education in coding and AI-related topics [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. However, acquiring such skills remains challenging due to intensive curricular demands and limited exposure to computational concepts.\u003c/p\u003e \u003cp\u003eMedical students face several specific challenges in AI-related programming education. These include difficulty understanding abstract programming logic without sufficient foundational training, limited opportunities for hands-on practice, and lack of real-time guidelines during coding exercises. Moreover, conventional instructional methods, often based on static lectures or textual assignments [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e], do not cater to individual learning needs and can lead to frustration or disengagement.\u003c/p\u003e \u003cp\u003eGenerative Artificial Intelligence (GenAI), particularly models based on large language models (LLMs), offers a promising solution to these educational challenges. Tools such as GitHub Copilot [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e] and ChatGPT [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e] enable natural language-based code generation and explanation, thereby lowering the learning curve and offering on-demand support. These technologies have been shown to alleviate programming anxiety, promote inquiry-based learning, and support students with diverse computational backgrounds by offering immediate, personalized feedback, and interactive problem-solving capabilities [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eDespite these advancements, limited research has examined how medical students perceive and interact with GenAI tools in the context of programming education. Previous studies have primarily compared GenAI capabilities with those of human experts, with little attention paid to learner-centered evaluation [\u003cspan additionalcitationids=\"CR12\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eTo address this gap, we developed Chat2R, a GenAI-powered programming assistant specifically designed to support medical students in acquiring programming and AI-related skills. Chat2R improves upon traditional instruction by offering medical students real-time, natural language-driven support in three critical areas: code generation, code explanation, and code debugging. Unlike conventional teaching methods, which often rely on static lecture content and delayed feedback, Chat2R enables students to interactively generate R code from plain language prompts, receive instant explanations to enhance their understanding, and iteratively debug code with AI assistance. These features make programming more accessible, adaptive, and student-centered. Furthermore, Chat2R accommodates varied learning paces, allowing students to progress according to their comfort level, thereby reducing anxiety and promoting confidence. We conducted a cross-sectional study using a mixed-methods approach to evaluate Chat2R\u0026rsquo;s effectiveness, usability, and impact on students\u0026rsquo; learning experience.\u003c/p\u003e"},{"header":"2 METHOD","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Participants\u003c/h2\u003e \u003cp\u003eThis cross-sectional study was conducted with postgraduate students enrolled in the elective course \"Practical Techniques of Medical Data Mining\" (Course No. INSC11011) at Peking Union Medical College (PUMC) [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. At the beginning of the semester, 50 students from various disciplines registered for the course. All enrolled students were invited to participate in the study, and participation was entirely voluntary. No exclusion criteria were applied. Ultimately, 31 students who attended the course and completed the questionnaire were included in the analysis. The other students either chose not to participate or did not use the GenAI-based programming assistance tool, Chat2R, and were therefore excluded from the final dataset.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Design of the programming assistance tool: Chat2R\u003c/h2\u003e \u003cp\u003eChat2R is a programming assistance tool that leverages generative artificial intelligence techniques to streamline the coding process for medical students. The tool assists with three primary functions: code generation, code explanation, and code debugging. It automates the creation of code based on user\u0026rsquo;s natural language inputs, reducing the time and effort required for students to write code from scratch. Additionally, it provides detailed explanations of generated code to help students understand the logic and structure behind the code, and it assists students in identifying and correcting errors in their code, thereby improving their problem-solving skills and understanding of programming concepts. The flowchart in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e illustrates the implementation and workflow of Chat2R, the code generation tool leveraging the Qwen-72B-Chat [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e] interface to assist in learning R programming. Chat2R starts by receiving a problem description and creating an initial prompt to generate R code. Configured through a system prompt, Chat2R produces the code, which the user then executes. If an error occurs during execution, Chat2R generates a new prompt to revise and improve the code. This iterative process of generating prompts, refining code, and user execution continues until the code achieves the expected results, effectively solving the given problem. The design of Chat2R focuses on user-friendliness and educational value, ensuring seamless integration into the medical data mining course. The tool is accessible via a web interface, allowing students to interact with it using natural language queries.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Design of the questionnaire\u003c/h2\u003e \u003cp\u003eA 26-item questionnaire was developed based on relevant literature and refined through consultation with two medical artificial intelligent education experts to ensure content validity [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e][\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. It comprised four sections: (1) demographic information (Q1-Q4); (2) knowledge and familiarity with GenAI (Q5-Q11); (3) perceptions and experiences using Chat2R (Q12-Q18); (4) perspectives on GenAI\u0026rsquo;s potential role in clinical practice (Q19-Q24). The questionnaire included multiple item types. Items Q5-Q11 and Q15-Q24 were measured using a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree), designed to assess students\u0026rsquo; familiarity with GenAI technologies and their attitudes toward their application in education and clinical practice. Other items (e.g., Q12-Q14) utilized frequency-based or binary response formats (e.g., Yes/No) to capture students\u0026rsquo; usage behaviors and overall impressions of Chat2R. The questionnaire also included two open-ended questions to capture quantitative feedback.\u003c/p\u003e \u003cp\u003eA pilot test with five postgraduate students was conducted to assess clarity, comprehensibility, and item coherence. Based on the feedback, minor revisions were made to improve wording and structure. To evaluate internal consistency, Cronbach\u0026rsquo;s alpha was calculated using the full dataset from the main study (n\u0026thinsp;=\u0026thinsp;31). The resulting alpha coefficient was 0.73, indicating acceptable reliability of the instrument.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4 Data collection process\u003c/h2\u003e \u003cp\u003eThe data collection process involved multiple stages to ensure comprehensive and reliable data. The questionnaire was administered to students\u0026rsquo; post-implementation of Chat2R, ensuring that they had sufficient time to interact with the tool. Instructors observed student interactions with Chat2R during the course, noting any challenges or areas where students required additional support. Detailed logs of student interactions with Chat2R were collected, capturing the types of queries and issues encountered.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.5 Analysis\u003c/h2\u003e \u003cp\u003eThe analysis of the collected data was conducted through a multi-stage approach to ensure a comprehensive evaluation of Chat2R\u0026rsquo;s effectiveness in medical education. Initially, descriptive statistics were applied to the survey responses to summarize students\u0026rsquo; perceptions regarding Chat2R's ease of use, educational value, and overall usefulness. To gain deeper insights into the qualitative data obtained from open-ended responses, thematic analysis was conducted, involving the systematic coding of responses to identify recurring themes and patterns related to students' experiences and challenges when using Chat2R.\u003c/p\u003e \u003cp\u003eThe quantitative and qualitative findings were then integrated to provide a holistic assessment of Chat2R\u0026rsquo;s impact, using cross-validation to compare survey results with qualitative insights and ensure consistency and reliability. Additionally, logs of student interactions with Chat2R were analyzed to classify the types of queries submitted (e.g., code generation, code explanation, code debugging) and to identify common issues faced by students. This mixed-methods approach allowed for a nuanced understanding of how Chat2R supports the learning process and highlighted areas for further refinement and development.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e2.6 Ethics\u003c/h2\u003e \u003cp\u003e The Ethics Committee of the Institute of Medical Information, Chinese Academy of Medical Sciences and PUMC, granted approval for this study (IMICAMS/01/20/HREC). All participants were informed that their responses would contribute to public-facing research, and written informed consent was obtained prior to participation. All procedures were conducted in strict adherence to the principles outlined in the Declaration of Helsinki.\u003c/p\u003e \u003c/div\u003e"},{"header":"3 RESULTS","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Participant Demographics\u003c/h2\u003e \u003cp\u003eA total of 31 postgraduate students completed the questionnaire. Participants represented 14 academic disciplines, with 18 females (58.06%) and 13 males (41.94%). The average age was 25 years (range: 20\u0026ndash;44). This diverse sample included students from fields such as medical informatics, internal medicine, public health, and nursing. Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e provides a detailed breakdown of these demographics, illustrating the multidisciplinary nature of the participants involved in the research.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eParticipants\u0026rsquo; demographic information\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN(%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e%\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSex\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e41.94\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e58.06\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMajor\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMedical Informatics\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e19.4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInternal Medicine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e12.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBiomedical Engineering\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e12.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInformation Science\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e9.7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEpidemiology and Health Statistics\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e9.7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eClinical Diagnostics\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e9.7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eImaging and Nuclear Medicine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGeriatrics\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePublic Health\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNursing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePediatrics\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eClinical Anesthesiology\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eClinical Integration of Traditional Chinese and Western Medicine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eClinical Medicine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Knowledge and familiarity with GenAI Technologies\u003c/h2\u003e \u003cp\u003eStudents reported a generally high level of awareness and understanding of GenAI technologies. The average composite score for Q5-Q11 was 4.32 (SD\u0026thinsp;=\u0026thinsp;0.48) on a 5-point Likert scale (1\u0026thinsp;=\u0026thinsp;strongly disagree, 5\u0026thinsp;=\u0026thinsp;strongly agree). The highest-rated item was Q7: \u0026ldquo;I understand generative AI technologies like ChatGPT can generate output that is factually inaccurate\u0026rdquo; (Mean\u0026thinsp;=\u0026thinsp;4.35, SD\u0026thinsp;=\u0026thinsp;0.55), while the lowest-rated item was Q9: \u0026ldquo;I understand generative AI technologies like ChatGPT can exhibit biases and unfairness in their output\" (Mean\u0026thinsp;=\u0026thinsp;3.65, SD\u0026thinsp;=\u0026thinsp;0.88).\u003c/p\u003e \u003cp\u003eStudents from technical disciplines such as epidemiology, medical informatics, and biomedical engineering scored higher on GenAI knowledge than those from clinical majors (see Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Chat2R Usage and Perceived Usefulness\u003c/h2\u003e \u003cp\u003eAmong the 31 students surveyed, 23 (74.2%) reported prior experience with GenAI products before the course (Q13), and an equal number (23, 74.2%) used Chat2R during the course. Chi-square analysis revealed no significant association between prior GenAI experience and actual usage of Chat2R during class sessions (χ\u0026sup2; (1, N\u0026thinsp;=\u0026thinsp;31)\u0026thinsp;=\u0026thinsp;0.0, p\u0026thinsp;\u0026gt;\u0026thinsp;0.05), suggesting that prior exposure to GenAI tools was not a determining factor for Chat2R engagement.\u003c/p\u003e \u003cp\u003eIn terms of usage frequency (Q14), over half of Chat2R users (13, 56.52%) reported using the tool weekly, while the remaining 43.48% used it occasionally. Most users found the tool easy to use (Q15): 8 students (34.78%) rated it as \u0026ldquo;very easy\u0026rdquo;, 14 students (60.87%) as \u0026ldquo;easy\u0026rdquo;, and 1 (4.35%) as \u0026ldquo;medium\u0026rdquo;. However, no significant association was found between usage frequency and perceived ease of use (χ\u0026sup2; (2, N\u0026thinsp;=\u0026thinsp;23)\u0026thinsp;=\u0026thinsp;2.65, p\u0026thinsp;\u0026gt;\u0026thinsp;0.05).\u003c/p\u003e \u003cp\u003eWhen asked whether Chat2R helped their learning (Q16), a large majority responded positively: 12 students (38.71%) rated it as \u0026ldquo;very helpful,\u0026rdquo; and 18 (58.06%) as \u0026ldquo;helpful,\u0026rdquo; with only one student selecting \u0026ldquo;medium.\u0026rdquo; Subgroup analysis showed that all 23 Chat2R users considered the tool either \u0026ldquo;very helpful\u0026rdquo; or \u0026ldquo;helpful.\u0026rdquo; Among non-users, most still perceived the tool as helpful based on peer observation or classroom context, though their ratings were slightly less enthusiastic.\u003c/p\u003e \u003cp\u003eStudents\u0026rsquo; overall impressions of Chat2R (Q17) were also favorable. 13 students (41.94%) rated the tool as \u0026ldquo;very good,\u0026rdquo; 16 (51.61%) as \u0026ldquo;good,\u0026rdquo; and 2 (6.45%) as \u0026ldquo;medium.\u0026rdquo; Interestingly, students who used Chat2R tended to have a more uniformly positive impression, while responses from non-users were more varied. A statistically significant difference in overall impressions between users and non-users was observed (χ\u0026sup2; test, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e \u003cp\u003eRegarding future use intentions (Q18), most students expressed willingness to continue using Chat2R. Specifically, 15 (48.39%) were \u0026ldquo;very willing,\u0026rdquo; 12 (38.71%) \u0026ldquo;willing,\u0026rdquo; 3 (9.68%) \u0026ldquo;neutral,\u0026rdquo; and only one (3.23%) was \u0026ldquo;not willing.\u0026rdquo; That single student who expressed unwillingness had not used Chat2R in class. However, there was no statistically significant association between prior usage and future willingness to use the tool (χ\u0026sup2; (3, N\u0026thinsp;=\u0026thinsp;31)\u0026thinsp;=\u0026thinsp;3.92, p\u0026thinsp;\u0026gt;\u0026thinsp;0.05).\u003c/p\u003e \u003cp\u003eOverall, these results suggest that Chat2R was perceived as a useful, accessible, and well-received learning tool by most students, including those with no prior coding or AI experience.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Perspectives on GenAI\u0026rsquo;s Potential Role in Clinical Practice\u003c/h2\u003e \u003cp\u003eStudents generally expressed strong support for the integration of GenAI into future clinical workflows. On the 5-point Likert scale, the highest-rated item was Q21: \u0026ldquo;\u003cem\u003eGenerative AI technologies can help me save time\u003c/em\u003e\u0026rdquo; (Mean\u0026thinsp;=\u0026thinsp;4.65, SD\u0026thinsp;=\u0026thinsp;0.49), indicating a strong belief in the efficiency-enhancing potential of these tools. The lowest-rated item was Q20: \u0026ldquo;\u003cem\u003eGenerative AI technologies can improve my computational thinking skills\u003c/em\u003e\u0026rdquo; (Mean\u0026thinsp;=\u0026thinsp;3.65, SD\u0026thinsp;=\u0026thinsp;0.66), suggesting some uncertainty about the long-term impact of GenAI on students\u0026rsquo; technical skill development.\u003c/p\u003e \u003cp\u003eStudents also expressed high agreement with statements reflecting the practical utility of GenAI tools: the ability to offer personalized and immediate feedback (Mean\u0026thinsp;=\u0026thinsp;4.23, SD\u0026thinsp;=\u0026thinsp;0.56), generate unique insights (Mean\u0026thinsp;=\u0026thinsp;4.32, SD\u0026thinsp;=\u0026thinsp;0.65), and provide around-the-clock availability (Mean\u0026thinsp;=\u0026thinsp;4.35, SD\u0026thinsp;=\u0026thinsp;0.75). These findings indicate that students perceive GenAI as a valuable supplementary assistant in clinical practice.\u003c/p\u003e \u003cp\u003eFurther analysis of willingness scores across different academic disciplines showed notable variation. Students majoring in Medical Informatics (Median\u0026thinsp;=\u0026thinsp;27) and Epidemiology and Health Statistics (Median\u0026thinsp;=\u0026thinsp;26) demonstrated the highest levels of willingness to adopt GenAI technologies in future practice. In contrast, students from Clinical Diagnostics reported the lowest median score (Median\u0026thinsp;=\u0026thinsp;22), as illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e3.5 Qualitative Findings\u003c/h2\u003e \u003cp\u003e Open-ended responses provided additional insights into students\u0026rsquo; perceptions of Chat2R. When asked about its role in supporting their learning, many students highlighted Chat2R\u0026rsquo;s utility in assisting with code correction and debugging. Several respondents noted that the tool helped them identify and resolve errors, clarify coding logic, and simplify complex expressions. In their words, Chat2R \u0026ldquo;\u003cem\u003ecorrected errors, explained code functions, and simplified logic\u003c/em\u003e\u0026rdquo; and \u0026ldquo;\u003cem\u003eprovided coding ideas and alternative approaches\u003c/em\u003e\u0026rdquo;. Beyond technical support, students also described Chat2R as facilitating clearer thinking and offering diverse perspectives when tackling programming tasks. For instance, one student remarked that the tool \u0026ldquo;\u003cem\u003ehelped structure thoughts and explore different problem-solving strategies\u003c/em\u003e\u0026rdquo;. In addition, Chat2R was perceived to enhance learning efficiency and promote independent exploration, with multiple students commenting on its ability to \u0026ldquo;\u003cem\u003eimprove speed and autonomy\u003c/em\u003e\u0026rdquo; when writing and modifying code.\u003c/p\u003e \u003cp\u003eStudents also offered suggestions for improving the tool and its implementation. Several participants expressed a desire for extended access to Chat2R after the course concluded, while others recommended incorporating more practical case-based examples and enabling file or image uploads to support real-world data analysis scenarios. Additional suggestions included improving the user interface, such as easier access to chat history, and expanding compatibility with other statistical tools. A few students also proposed exploring advanced AI topics, including deep learning, in future iterations of the tool. Collectively, these qualitative insights underscore Chat2R\u0026rsquo;s perceived effectiveness as a supportive educational tool while also highlighting areas for refinement to better align with learners\u0026rsquo; evolving needs.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e3.6 Interaction Behavior Analysis in Chat2R\u003c/h2\u003e \u003cp\u003eTo better understand how students engaged with the Chat2R tool, interaction logs were analyzed with a focus on session structure, query patterns, content relevance, and input characteristics.\u003c/p\u003e \u003cp\u003eAs shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e, most chat sessions consisted of a single round of interaction, indicating that most students used Chat2R to obtain quick answers rather than to engage in extended, multi-turn conversation. This suggested that the tool was primarily used for task-specific assistance. To further characterize usage patterns, the proportion of follow-up versus new questions was examined. As illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e, follow-up questions occurred more frequently than entirely new prompts. This suggests an iterative learning behavior in which students used Chat2R to refine previous outputs or clarify ongoing tasks, rather than starting with unrelated queries each time.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eNext, the questions were categorized by their relevance to different stages of the data analysis process, including data collection, preprocessing, exploration, visualization, and modeling. Of the 1,603 questions submitted, 311 could be directly mapped to a defined analysis phase: 138 related to data collection, 74 to preprocessing, 4 to exploration, 87 to visualization, and 8 to modeling (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e). The remaining questions focused primarily on general code explanation, debugging, and syntax clarification, and thus did not correspond to a single analytical stage.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFinally, we examined the distribution of question lengths to better understand students\u0026rsquo; prompting behavior and potential reliance on the tool. As shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e, most queries were relatively brief-1,457 of 1,603 questions contained fewer than 200 characters. The most frequent input length was 11 characters (n\u0026thinsp;=\u0026thinsp;74), with question lengths ranging from 1 to 1,375 characters. These findings indicate that students tend to use short, targeted queries to interact with Chat2R, consistent with task-oriented and efficiency-driven usage.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"4 DISCUSSIONS","content":"\u003cp\u003eThis research systematically evaluated the usability and perceived effectiveness of large language models (LLMs), specifically Chat2R, as an assistance tool for programming education among medical students. A mixed-methods approach was adopted, combining quantitative survey results with qualitative insights to assess both user experience and the pedagogical value of GenAI technologies. The findings demonstrated that Chat2R provided practical benefits in the context of medical programming education. Students reported that the tool facilitated learning through real-time code generation, explanation, and debugging, which aligns with previous research suggesting that AI-powered educational technologies can increase learner engagement, lower cognitive burden, and promote programming confidence [\u003cspan additionalcitationids=\"CR18\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. These results are consistent with studies indicating that immediate feedback and AI-enabled guidelines improve student autonomy and reduce frustration in learning complex skills [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eBy incorporating Chat2R into courses that focus on data analysis and medical programming, educational institutions can facilitate interactive, student-centered learning that helps students overcome common obstacles in programming. This integration could mitigate the frustration often associated with traditional programming instruction [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. Additionally, implementing Chat2R in flipped classroom models, where students use the tool for coding practice at home, would allow in-class time to be devoted to deeper learning through collaborative problem-solving and discussion.\u003c/p\u003e \u003cp\u003eChat2R\u0026rsquo;s real-time coding assistance can be extended beyond the classroom to include broader student support systems, such as coding labs and tutoring services. This broader implementation would offer students additional resources for independent learning and promote the development of problem-solving skills essential for medical practice. Studies on supplemental learning resources have shown that access to support tools outside of classroom hours contributes to a deeper understanding of course material and improves academic performance [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Deploying Chat2R in supplementary learning contexts could provide personalized feedback aligned with students\u0026rsquo; varying levels of proficiency. Research on adaptive learning technologies highlights that tailored support enhances learning retention and student satisfaction [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. The tool's capability to cater to different learning paces means that students requiring more time can receive targeted help, while advanced learners can explore complex concepts at their own speed. This flexibility addresses the diversity of programming experience among medical students, supporting them in building foundational and advanced skills that are increasingly essential in medical research and practice.\u003c/p\u003e \u003cp\u003eDespite its advantages, concerns remain regarding the potential for over-reliance on Chat2R. This aligns with earlier studies cautioning that excessive dependence on AI-based educational tools may foster surface-level learning and hinder the development of independent problem-solving skills [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e][\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. To mitigate such risks, it is essential that GenAI tools be integrated within structured and pedagogically sound learning environments that emphasize guided instruction, critical reflection, and active engagement with the learning tasks.\u003c/p\u003e \u003cp\u003eEffective implementation of Chat2R and similar tools also depends on targeted faculty development. Instructors require training not only in the technical use of GenAI platforms but also in evidence-based strategies for facilitating productive student-AI interaction [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. Prior research has demonstrated that professional development in educational technology enhances instructors\u0026rsquo; capacity to design adaptive, inclusive, and student-centered learning environments [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. Institutional support is equally important to ensure that GenAI integration aligns with curricular goals, adheres to ethical standards, and maintains technical reliability.\u003c/p\u003e \u003cp\u003eThis study has several limitations that must be acknowledged. Most notably, the absence of a control group limits the ability to draw causal inferences. Although some students did not actively use Chat2R, the study was not designed to compare their outcomes systematically with those of users. Furthermore, reliance on self-reported data introduces potential biases, including recall inaccuracies and social desirability effects. Future research should address these limitations by employing randomized controlled trials or longitudinal study designs, enabling the tracking of learning outcomes and perceptions over time. Incorporating objective performance indicators, such as coding assessments or log-based learning analytics, would provide a more comprehensive evaluation of Chat2R\u0026rsquo;s educational impact.\u003c/p\u003e \u003cp\u003eLooking ahead, the functionality of Chat2R could be further expanded through the integration of AutoML [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e](automated machine learning) capabilities. This enhancement would allow students to transition from manual coding tasks to higher-order activities such as model interpretation and clinical insights generation. Such development aligns with broader trends in which AI augments, rather than replaces human expertise in health data analysis [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. By enabling learners to move from implementation to strategic oversight, Chat2R has the potential to prepare future medical professionals to critically supervise AI-driver tools in both research and clinical contexts.\u003c/p\u003e \u003cp\u003eIn summary, Chat2R demonstrates strong potential as a generative AI\u0026ndash;powered educational tool for supporting programming education among medical students. With thoughtful instructional design, professional development, and institutional support, such tools can become integral components of modern medical education, fostering more personalized, interactive, and effective learning environments.\u003c/p\u003e"},{"header":"5 CONCLUSIONS","content":"\u003cp\u003eThis study demonstrates that integrating Chat2R, a generative AI-based code generation tool, into the \"Practical Techniques of Medical Data Mining\" course effectively supported medical students in acquiring programming skills. Students reported a positive learning experience, noting enhanced computational thinking and reduced barriers to engaging with coding tasks. Analysis of usage patterns revealed diverse learning needs, with a strong emphasis on code generation and moderate reliance on debugging support. These findings suggest that GenAI tools such as Chat2R have the potential to improve the accessibility and personalization of programming education in medical training. Further research should explore the long-term effects of such technologies on learning outcomes across diverse educational settings.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eConflict of Interest Statement\u003c/h2\u003e \u003cp\u003eThe authors declare that they have no conflict of interest.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eXiaowei Xu: Methodology, Software, Data curation, Investigation, Writing - original draft. Zheng Si: Validation, Investigation. Lin Yang: Validation. Jie Hao: Methodology. Xuwen Wang: Data curation. Hongyu Kang: Validation. Chao Ma: Supervision. Xudong Lv: Supervision. Jiao Li: Funding acquisition, Conceptualization, Supervision. All authors contributed to the development of this manuscript and approved the final version for submission.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eWe gratefully appreciate all of the participants and staff for their contributions.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe datasets generated during the current study are not publicly available due to participant confidentiality and institutional restrictions, but are available from the corresponding author on reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAhsan MM, Luna SA, Siddique Z. Machine-learning-based disease diagnosis: a comprehensive review. Healthcare. 2022;10(3):541. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/healthcare10030541\u003c/span\u003e\u003cspan address=\"10.3390/healthcare10030541\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShafi S, Parwani AV. Artificial intelligence in diagnostic pathology. Diagn Pathol. 2023;18(1):109. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s13000-023-01375-z\u003c/span\u003e\u003cspan address=\"10.1186/s13000-023-01375-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang Z, Yang E, Shen J et al. A pathologist-AI collaboration framework for enhancing diagnostic accuracies and efficiencies. Nat Biomed Eng. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41551-024-01223-5\u003c/span\u003e\u003cspan address=\"10.1038/s41551-024-01223-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGala D, Behl H, Shah M, Makaryus AN. The Role of Artificial Intelligence in Improving Patient Outcomes and Future of Healthcare Delivery in Cardiology: A Narrative Review of the Literature. Healthc (Basel). 2024;12(4):481. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/healthcare12040481\u003c/span\u003e\u003cspan address=\"10.3390/healthcare12040481\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarone L, Williams J, Micklos D. Unmet needs for analyzing biological big data: A survey of 704 NSF principal investigators. PLoS Comput Biol. 2017;13(10):e1005755. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1371/journal.pcbi.1005755\u003c/span\u003e\u003cspan address=\"10.1371/journal.pcbi.1005755\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStanford Medicine 2020 Health Trends Report. The Rise of the Data-Driven Physician. Stanford, CA: Stanford Medicine URL: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://med.stanford.edu/content/dam/sm/school/documents/Health-Trends-Report\u003c/span\u003e\u003cspan address=\"https://med.stanford.edu/content/dam/sm/school/documents/Health-Trends-Report\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e/Stanford Medicine Health Trends Report 2020.pdf [accessed 2024-09-16].\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang L, Zheng S, Xu X, et al. Medical Data Mining Course Development in Postgraduate Medical Education: Web-Based Survey and Case Study. JMIR Med Educ. 2021;7(4):e24027. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2196/24027\u003c/span\u003e\u003cspan address=\"10.2196/24027\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGitHub Copilot. URL: https://github.com/features/copilot [accessed 2024-11-16].\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChatGPT. URL:https://chat.openai.com/ [accessed 2024-11-16].\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTayebi Arasteh S, Han T, Lotfinia M, et al. Large language models streamline automated machine learning for clinical studies. Nat Commun. 2024;15(1):1603. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41467-024-45879-8\u003c/span\u003e\u003cspan address=\"10.1038/s41467-024-45879-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFan A, Gokkaya B, Harman M et al. Large language models for software engineering: Survey and open problems. In: 2023 IEEE/ACM International Conference on Software Engineering: Future of Software Engineering (ICSE-FoSE); 2023:31\u0026ndash;53. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1109/ICSE-FoSE59343.2023.00008\u003c/span\u003e\u003cspan address=\"10.1109/ICSE-FoSE59343.2023.00008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChang Y, Wang X, Wang J et al. A survey on evaluation of large language models. ACM Trans Intell Syst Technol. 2024;15(3): Article 39. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1145/3641289\u003c/span\u003e\u003cspan address=\"10.1145/3641289\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKang K, Yang Y, Wu Y, Luo R. Integrating Large Language Models in Bioinformatics Education for Medical Students: Opportunities and Challenges. Ann Biomed Eng. 2024;52(9):2311\u0026ndash;5. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s10439-024-03554-5\u003c/span\u003e\u003cspan address=\"10.1007/s10439-024-03554-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChinese Academy of Medical Sciences and Peking Union Medical College. URL: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://english.cams.cn/index.html\u003c/span\u003e\u003cspan address=\"http://english.cams.cn/index.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e [accessed 2024-11-16].\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBai J, Bai S, Chu Y et al. Qwen technical report. arXiv. Preprint posted online September 28, 2023. Arxiv: 2309.16609.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMai D, Da C, Hanh N. The use of ChatGPT in teaching and learning: a systematic review through SWOT analysis approach. Front Educ. 2024;9:1328769. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/feduc.2024.1328769\u003c/span\u003e\u003cspan address=\"10.3389/feduc.2024.1328769\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChan C, Hu W. Students\u0026rsquo; voices on generative AI: perceptions, benefits, and challenges in higher education. Int J Educ Technol High Educ. 2023;20:43. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s41239-023-00411-8\u003c/span\u003e\u003cspan address=\"10.1186/s41239-023-00411-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHooda M, Rana C, Dahiya O et al. Artificial Intelligence for Assessment and Feedback to Enhance Student Success in Higher Education. Math Probl Eng. 2022; 2022: 1\u0026ndash;19. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1155/2022/5215722\u003c/span\u003e\u003cspan address=\"10.1155/2022/5215722\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLabadze L, Grigolia M, Machaidze L. Role of AI chatbots in education: systematic literature review. Int J Educ Technol High Educ. 2023;20:56. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s41239-023-00426-1\u003c/span\u003e\u003cspan address=\"10.1186/s41239-023-00426-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTaylor D, Yeung M, Bashet A. Personalized and Adaptive Learning. In: Ryoo J, Winkelmann K, editors. Innovative Learning Environments in STEM Higher Education. Springer Briefs in Statistics. Cham: Springer; 2021. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/978-3-030-58948-6_2\u003c/span\u003e\u003cspan address=\"10.1007/978-3-030-58948-6_2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWeng X, Chiu T. Instructional design and learning outcomes of intelligent computer assisted language learning: Systematic review in the field. Comput Educ Artif Intell. 2023;4:100117. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.caeai.2022.100117\u003c/span\u003e\u003cspan address=\"10.1016/j.caeai.2022.100117\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi Z, Chen H, Gao X. The influencing factors and predictability of primary school students\u0026rsquo; learning performance in online supplementary classes. Educ Inf Technol. 2024;29:10995\u0026ndash;1021. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s10639-023-12246-x\u003c/span\u003e\u003cspan address=\"10.1007/s10639-023-12246-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGerlich M. AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies. 2025;15(1):6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/soc15010006\u003c/span\u003e\u003cspan address=\"10.3390/soc15010006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee H, Lee HP, Sarkar A, Tankelevitch L et al. The impact of generative AI on critical thinking: self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. Proc ACM CHI Conf Hum Factors Comput Syst. 2025.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMulyani H, Istiaq M, Shauki E, et al. Transforming education: exploring the influence of generative AI on teaching performance. Cogent Educ. 2025;12(1). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1080/2331186X.2024.2448066\u003c/span\u003e\u003cspan address=\"10.1080/2331186X.2024.2448066\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUzorka A, Namara S, Olaniyan A. Modern technology adoption and professional development of lecturers. Educ Inf Technol. 2023;28:14693\u0026ndash;719. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s10639-023-11790-w\u003c/span\u003e\u003cspan address=\"10.1007/s10639-023-11790-w\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang L, Basham J, Carter R, et al. Exploring Factors associated with the implementation of student-centered instructional practices in U.S. classroom. Teach Teach Educ. 2021;99:103273. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.tate.2020.103273\u003c/span\u003e\u003cspan address=\"10.1016/j.tate.2020.103273\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAutoML.org. AutoML: Automated Machine Learning. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.automl.org/automl/\u003c/span\u003e\u003cspan address=\"https://www.automl.org/automl/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. [accessed 2024-11-16].\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYu K, Healey E, Leong T, et al. Medical Artificial Intelligence and Human Values. N Engl J Med. 2024;390(20):1895\u0026ndash;904. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1056/NEJMra2214183\u003c/span\u003e\u003cspan address=\"10.1056/NEJMra2214183\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-education","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"meed","sideBox":"Learn more about [BMC Medical Education](http://bmcmededuc.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/meed/default.aspx","title":"BMC Medical Education","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Generative AI, student perception, code generation, programming education","lastPublishedDoi":"10.21203/rs.3.rs-5606661/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5606661/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackgroud\u003c/h2\u003e \u003cp\u003eWith the increasing importance of computational skills in healthcare, there is a growing need to equip medical students with programming knowledge to address complex healthcare challenges effectively. Traditional programming methods, however, can be challenging for medical students due to heavy academic loads and limited exposure to coding. Generative artificial intelligence (GenAI) presents a promising solution to these issues.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThis research evaluated the feasibility of Chat2R, an R programming assistance tool powered by GenAI, within the postgraduate course \u0026ldquo;Practical Techniques of Medical Data Mining\u0026rdquo; at Peking Union Medical College (PUMC). A mixed-methods approach was used, combining quantitative surveys and qualitative insights to assess the tool\u0026rsquo;s effectiveness and students\u0026rsquo; reception. Quantitative data was gathered through post-implementation surveys measuring students' perceptions of their coding proficiency and the tool\u0026rsquo;s utility. Qualitative analysis explored student interactions with Chat2R, identifying key challenges and concerns to enhance the educational experience.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eA total of 31 postgraduate students from 14 different disciplines participated in the survey. The positive feedback supported the integration of Chat2R as a valuable educational tool, highlighting GenAI\u0026rsquo;s role in enhancing computational thinking skills. Between March 13 and April 2, 2024, 28 students actively engaged with Chat2R, generating 1,603 questions. Of these, 311 were specifically related to defined data analysis processes: 138 questions on data collection, 74 on data preprocessing, 4 on data exploration, 87 on data visualization, and 8 to data modeling.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThis study demonstrates that integrating Chat2R, a GenAI-based programming tool, can significantly enhance programming education for medical students. It improves computational thinking, supports both lecture and practical learning, and addresses challenges related to limited coding exposure. Positive student feedback highlights its effectiveness in providing coding assistance and fostering an interactive, student-centered learning environment. The findings also underscore the importance of professional development for educators to effectively incorporate GenAI tools into teaching. Future enhancements could include AutoML capabilities, enabling medical students to guide AI-driven data analysis and better utilize AI in clinical and research contexts.\u003c/p\u003e","manuscriptTitle":"Integrating Generative AI-Based Assistance Tool in Programming Education for Medical Students: A Cross-Sectional Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-14 22:57:52","doi":"10.21203/rs.3.rs-5606661/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-04-21T11:34:57+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-04-21T10:52:03+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-04-13T06:25:14+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"7433490278183460864927243670345480454","date":"2025-04-12T05:30:07+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"314417376236991276278527018829601453886","date":"2025-04-10T06:17:46+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-04-09T18:57:31+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-04-07T10:45:57+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Education","date":"2025-04-07T05:25:08+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-education","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"meed","sideBox":"Learn more about [BMC Medical Education](http://bmcmededuc.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/meed/default.aspx","title":"BMC Medical Education","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"340ddc5a-98b6-4d25-933a-85d6f18dd5f7","owner":[],"postedDate":"April 14th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-07-07T01:53:08+00:00","versionOfRecord":[],"versionCreatedAt":"2025-04-14 22:57:52","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5606661","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5606661","identity":"rs-5606661","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}