Challenges and opportunities for teaching and learning in an era of generative artificial intelligence: Students’ perspective

INTRODUCTION

Generative artificial intelligence (GenAI) offers a transformative leap beyond traditional educational technologies (EdTech) and conventional artificial intelligence (AI) by dynamically creating original, context-specific content in response to the user. The traditional AI analyses existing data, while GenAI creates new content through learning from it. Traditional AI focuses on categorisation, prediction and pattern recognition, while GenAI learns from data to produce original outputs, such as literature, graphics, or music.^[1,2] Literature suggests that GenAI is reshaping medical education by making it more interactive and efficient, potentially leading to better-prepared health professionals.

GenAI creates many opportunities for enhancing personalised learning experiences among students by assessing their learning preferences, styles and performance statistics, thereby providing tailored activities and resources.^[1,2] The shift from a one-size-fits-all approach to a more personalised approach allows students to progress at their own pace, enhancing their comprehension and retention of the material.^[2,3] GenAI can democratise education by providing quality resources to marginalised regions and diverse learners, generating content in multiple languages and media forms, thereby enhancing learning accessibility.^[3,4] This confirms support for learners with special needs and customises their learning approach to fit their feet.^[4] The tool can enhance learner participation and interactivity in group tasks, discussions and brainstorming sessions, transcending boundaries to cultivate diverse ideas and perspectives.^[3,5] AI tool helps students identify knowledge gaps, create reading materials, brainstorm, encourage group discovery and provide emotional support, sparking meaningful educational conversations.^[6,7] Furthermore, GenAI can enhance administrative efficiencies for educators, such as lesson planning and grading, allowing them to concentrate on teaching activities and mentoring.^[5] This may benefit learners by enabling them to connect with the educator regarding time, thus supporting them both academically and personally.^[5,8] However, it may lead to increased cost of education or problems of inaccessibility in resource-constrained areas and communities.

Researchers highlight challenges in the use of GenAI, including ethical concerns about data privacy and reliability, as unfiltered datasets may produce biased or culturally insensitive educational resources.^[4,7,8] Excessive dependence on GenAI for task completion can hamper the learner’s critical thinking and problem-solving ability, which, in turn, can hamper the quick decision-making process in those students.^[7,9] The idea is to complement the old-school teaching strategy rather than replace it to ensure the best of both teaching-learning approaches.^[2,8,10] Students must understand that while these tools can enhance learning, they must be used responsibly to prevent potential misuse.^[7,11] Addressing the digital divide in education is crucial due to disparities in access and opportunities, as well as resistance from educators and learners to adopt a growth mindset.^[2,5,12]

Keeping this background in mind, we explored medical students’ perceptions of the challenges and opportunities they face in the era of GenAI within educational settings in medical institutions in India. The research question was, ‘What is medical students’ perception regarding the usage of GenAI in the teaching-learning part of medical education?’ This approach will enable the medical faculty to enhance their teaching, learning and assessment strategies by utilising GenAI for improved student learning outcomes. Simultaneously, try to explore the challenges perceived and identify ways to address them.

MATERIALS AND METHODS

A cross-sectional survey was conducted using a structured questionnaire with 15 closed-ended and two open-ended questions. The questionnaire was developed through a literature review and the author’s expertise. The face validity of the perception questionnaire was assessed. Three experts in the field of medical education were provided with the study’s aim and objectives, as well as the developed questionnaire. The modifications and suggestions they supplied were discussed among the authors to finalise the survey questionnaire. The Institutional Ethics Committee approval was obtained (PUIECHR/PIMSR/00/081734/7806; Dated: 18 October 2024) before initiating data collection. The target population for this study includes undergraduate medical students in India. The invitation was forwarded to students from different medical colleges nationwide, comprising 8 medical colleges (4 private and 4 government) with approximately 5,600 enrolled undergraduate medical students. Three attempts were made to invite participants. The survey was administered through a Google Forms link to ensure accessibility and a broader reach. The link was disseminated through various online platforms, including email, university learning management systems and social media. The form/questionnaire description specified that filling out this form would be voluntary and that the data collected would be used for scholarly purposes. Once the participant submits the form, it is considered their consent. As a part of the study design, the authors have not asked the question regarding financial status to reduce hesitation in filling out the forms and to improve better acceptability of the participants to be part of the study. In addition, we steered clear of any bias stemming from social status. The minimum duration for participants to submit the questionnaire was 3 minutes. The authors ensured the confidentiality of the data.

The survey questionnaire contains demographics (Including age, gender, academic level and field of study), usage of AI tools (frequency and types of AI tools students used for scholarly work), perceived opportunities (perceived benefits of using AI for learning like personalised learning, enhanced creativity) and challenges (like over-reliance on AI, academic dishonesty and difficulty in critical thinking) and ethical concerns (plagiarism and fairness).

Statistical analysis was done using the Statistical Package for the Social Sciences-20. Descriptive statistics (mean, frequency and percentage) were used to summarise demographic data and students’ overall perceptions of AI. Inferential statistics (e.g. t-tests and Friedman’s ranking test) were conducted to identify significant differences in perceptions based on demographic variables. Reliability analysis, factor analysis and regression analysis were performed on the 15 closed-ended questions. Thematic analysis was used for open-ended responses.^[13] Two authors independently identified the codes and developed sub-themes and themes for the open-ended questions. All the authors collaborated to approve the themes and the sub-themes.

RESULTS

The final data from 173 participants (102 females and 71 males) were analysed out of 181 responses to the questionnaire, after checking for completeness, outliers and duplicate responses. All the students were undergraduate students pursuing MBBS degrees in various institutions across India. Students who participated from Government institutions numbered 99, and 74 from private institutions [Table 1]. Although we attempted to gather students’ perceptions of the undergraduate medical program across all academic levels, there was less participation from third- and 4^th-year students compared to first- and 2^nd-year students. Only 1^st- and 2^nd-year students from private medical colleges participated in the study. The comparison between government and private medical colleges could not be made as the degree of freedom for government participants was 3, while for private participants, it was 1.

The survey questionnaire shows a Cronbach’s Alpha reliability value of 0.879 for 15 items. The Cronbach’s Alpha for deleted items individually ranges from 0.864 to 0.889. Most participants have agreement responses for all items of the survey questionnaire related to GenAI in medical education, except for item P-8, which receives neutral responses from the participants [Figure 1]. Responses from medical undergraduate students as per gender and institute showed no statistical significance observed for any items based on gender, type of institute or year of program [Figure 2] except item P-5 for gender (0.037) and item P-9 for type of institute (P = 0.024). Based on the weighted mean calculated for the items (3.71), eight items showed high perception, while seven showed low perception. The Chi-square value (141.56) for 173 participants for the items shows statistical significance (P = 0.001) as per the Friedman mean rank test, with the highest rank for item P-3 ‘The use of GenAI in medical education can help simulate complex medical scenarios effectively’. (9.56) and lowest rank for the item P-8 ‘I feel confident in navigating the evolving landscape of AI in healthcare as a medical student’ (6.41) [Table 2]. The Kendall’s W^a of 0.06 with P < 0.001 showed that there is a low level of agreement, yet it is statistically significant. Post hoc Dunn showed no significant differences between the ranks of the items. No statistical significance was observed for mean perception and mean behaviour, as per gender, type of institute or year of medical program, among the participants.

Close

Figure 1:

The spectrum of responses from 1 to 5 for the survey item is represented by varied colour hues as per the frequency distribution of the responses by the undergraduate medical students. The number represents the frequency of the students for each item as per the Likert scale (1 = strongly disagree, 2 = disagree, 3 = neutral, 4 = agree and 5 = strongly agree).

Export to PPT

Close

Figure 2:

The graphical representation of undergraduate medical students’ responses to the survey questionnaire by gender and institute. P means items related to perception, while B relates to the behaviour of undergraduate medical students for the generative artificial intelligence.

Export to PPT

The factor analysis of the survey questionnaire yields a Kaiser-Meyer-Olkin (KMO) of 0.867 and a Bartlett’s test of sphericity that is statistically significant, with a Chi-square value of 1069.489. The component matrix utilises principal component analysis for factor analysis of the questionnaire, revealing four components extracted. However, when comparing them, the authors could only categorise them into two components (Perception Item 1–12; Behaviours Item 13–15).

Descriptive statistics for regression analysis show mean ± standard deviation values for mean perception and mean behaviour as 3.68 ± 0.56 and 3.85 ± 0.78, respectively. With mean perception as the independent variable and mean behaviour as the dependent variable, perception has a significant effect on behavioural attitude (F (1, 171) = 225.019, P < 0.001). The P-P plot of the regression standardised residuals also suggests a positive relationship between them [Figure 3]. The Pearson Correlation shows a strong correlation between the dependent and independent variables (r = 0.754). The Durbin-Watson value recorded was between 1 and 3 (1.768), showing that the data were independent of observation. The model is statistically significant (P < 0.01), indicating a strong correlation between behaviour and perception of participants for the GenAI.

Close

Figure 3:

The P-P plot of the regression standardised residual, suggesting a positive regression analysis between the mean_behaviour as the dependent and mean_perception as the independent variable.

Export to PPT

The thematic analysis was done for the open-ended questions. There were two questions; the first asked about the most important role of GenAI in the teaching-learning process and identified five significant themes, which were further categorised into 2–3 subthemes based on the participants’ responses. Whereas the other question addressed the challenges of using GenAI in the teaching-learning process, it identified six major themes, each with 2–3 subthemes [Table 3]. Further, suggestions provided by the participants for the use and challenges of GenAI in the teaching-learning process are categorised as Ethical and Professional Considerations, Integration and Implementation Strategies and Balancing AI and Human Expertise [Table 3].

DISCUSSION

The present study was an attempt to highlight the nuances of attitude among Indian medical students about the integration of GenAI in medical education. Today’s students frequently use technology for entertainment, social interactions and research. However, its integration into academia and teaching-learning must be strengthened, particularly in the health sciences. While digital technologies hold great promise for enhancing learning, their success depends largely on proper guidance toward structured usage and striking a balance between these innovations and more traditional teaching methods. Our observations align with the same; the participants opine that the use of GenAI enhances resource accessibility, visualisation and personalised learning in medical education and clinical training. It may improve efficiency by supporting diagnostic decision-making. However, there are social, cultural and psychological barriers to adoption and fear of reduced human interaction with ethical concerns. While students express strong endorsement of GenAI’s potential – for instance, ranking highly its capacity to simulate complex clinical scenarios (item P-3) – there remains a notable and seemingly contradictory lack of confidence in their ability to navigate the evolving AI landscape in healthcare (item P-8). This juxtaposition of enthusiastic support coupled with personal uncertainty suggests an important experiential and educational gap that warrants critical reflection. Rather than a simple acceptance or rejection, the findings reveal an ambivalent stance shaped by both optimism and perceived unpreparedness.

The developed questionnaire was shared with medical students from eight medical institutions across the country, receiving input from 181 participants from both government and private institutes. Only responses to P5 (between genders) and P9 (between types of institutions) were significantly different. The average scores agreed with using GenAI in teaching and learning in medical schools. The factor analysis for the questionnaire categorised the items into two major groups: perception (P-1 to P-12) and behaviour (B-1 to B-3). The output is divided into two major headings: Benefits and challenges.

CONCLUSION

The present study highlights that Gen Z medical students in India largely perceive GenAI as a transformative force in medical education. Students recognise the potential of GenAI in fostering personalised learning, facilitating comprehension of complex clinical scenarios and enhancing both teaching and patient care outcomes. They also appreciate its role in advancing scientific research and promoting efficiency and productivity within academic and clinical settings. However, despite this optimism, students expressed reservations regarding their own readiness to navigate the evolving landscape of AI in healthcare and raised important concerns relating to patient privacy, data security and the necessity of maintaining the core humanistic values within the medical profession.

Our findings underscore the urgent need for systematic and structured training modules on GenAI within the undergraduate medical curriculum. Such modules should not only enhance technical proficiency and critical thinking skills but also address ethical, legal and practical limitations associated with AI integration. The National Medical Commission and medical institutions must prioritise continuous faculty and student development programmes, invest in robust digital infrastructure and foster experiential learning to bridge the current preparedness gap.

In summary, meaningful and responsible integration of GenAI into Indian medical education will require a balanced approach – one that develops technological competence, safeguards ethical standards and upholds the principles of patient-centred care. Further multicentric research is warranted to evaluate the long-term impact of such curricular innovations and ensure sustainable utilisation of GenAI in medical education.