Learning Alongside AI: What Nursing Education Can Teach the Health Professions
Authors: Prof Jed Montayre1,2, Chun Hei Glen Cheng2
1. JBI Hong Kong Centre of Evidence-based Healthcare Excellence
2. The Hong Kong Polytechnic University
Artificial intelligence is becoming part of the typical milieu of healthcare education. It influences how students find and interpret information, how educators design learning activities and how future health professionals prepare for increasingly technology-enabled clinical environments. Among its most visible applications are generative AI chatbots, which can explain concepts, summarise material, generate clinical scenarios and provide conversational learning support.
The question, therefore, is no longer whether students will encounter AI. They already do. A more important question is how they can learn to use it responsibly without allowing convenience to replace curiosity, judgement or intellectual independence.
A systematic review and text network analysis that Jed co-authored examined the emerging literature on ChatGPT in nursing education. The study found that scholarly discussion was expanding rapidly but remained uneven. Much of the literature expressed optimism about ChatGPT’s educational potential, while also identifying significant ethical, academic and clinical concerns. Four themes were especially prominent: academic writing, healthcare simulation, data modelling and personal development.
Academic writing: Support or substitution?
Academic writing formed the largest thematic cluster. This is unsurprising because writing is often where students first encounter generative AI. A student can ask a chatbot to explain a difficult concept, suggest an outline, summarise an article or improve the clarity of a paragraph within seconds.
Used appropriately, these functions may help students overcome the anxiety of beginning an assignment and communicate their ideas more clearly. They may be particularly valuable for students writing in an additional language or those who benefit from immediate formative feedback.
However, the distinction between assistance and substitution is essential. If students use AI to organise ideas they have already developed, the technology may support learning. If they use it to avoid reading, analysis and reflection, it may produce polished work without meaningful understanding. The result can be an illusion of competence: the assignment appears complete, but the intellectual work that should have produced it has not occurred.
This matters in healthcare education because writing is not merely a method of assessment. It is also a means of learning how to reason. When students compare evidence, defend a clinical position and explain uncertainty, they are practising habits that later inform patient care. An efficiently generated answer cannot compensate for an underdeveloped capacity to think.
Healthcare simulation: Practising judgement before practice
A second theme concerned the use of ChatGPT in healthcare simulation. Generative AI can assist educators in developing patient scenarios, case studies, communication exercises and prompts that require students to make clinical decisions. It may also allow cases to be adapted to different levels of experience or learning objectives.
This offers genuine educational possibilities. Students could practise taking a history, communicating with a distressed patient, recognising changes in a clinical situation or explaining a care plan. Carefully designed AI-supported simulations may create additional opportunities for practice before students assume responsibility in real clinical settings.
Yet realism should not be confused with reliability. A scenario may sound clinically plausible while containing inaccurate, incomplete or inappropriate information. AI-generated simulations, therefore, require expert review and clearly defined learning outcomes. Their value depends less on the novelty of the technology than on the quality of the educational design surrounding it.
Data modelling: The hidden foundations of AI
Although data modelling was a smaller theme, it raised some of the most consequential questions. AI systems are shaped by their training data and by the assumptions embedded in their design. If those data are incomplete, outdated, unrepresentative or biased, the resulting outputs may reproduce the same limitations.
In healthcare, this is not simply a technical concern. It is an ethical one. An answer that overlooks cultural context, underrepresents certain populations or treats a general recommendation as universally applicable may contribute to inequitable care. Students should, therefore, understand that AI outputs are neither neutral nor inherently objective. They reflect the information available to the system, including its omissions and distortions.
AI literacy must consequently include more than knowing how to write an effective prompt. It requires asking whose knowledge is represented, whose experiences may be absent and how an apparently authoritative answer might affect different patients.
Personal development: Learning to question the answer
The final theme concerned students’ personal and professional development. Preparing students for an AI-enabled healthcare system requires confidence and technical competence, but it also requires intellectual humility, ethical awareness and critical judgement.
Students need to recognise that a fluent response is not necessarily an accurate one. They must be able to verify claims, examine the quality of cited evidence, identify uncertainty and recognise when a question exceeds the system’s capabilities. Most importantly, they need the confidence to disagree with an AI-generated answer.
The review’s sentiment analysis reflected this tension. Approximately 48% of the sentiments identified in the literature were positive, 31% were negative and 21% were neutral. These findings suggest cautious optimism rather than uncritical acceptance. Positive discussions emphasised opportunities for learning, writing, simulation and professional development. Negative discussions focused on misinformation, fabricated references, privacy, bias, academic misconduct and the possible erosion of human interaction. Neutral perspectives often reflected uncertainty and the need for stronger evidence, clearer institutional guidance and practical examples of responsible use.
Healthcare remains a human practice
During a talk in the Philippines, a student asked Jed a deceptively simple question: ‘If AI is so great, why don’t we just replace all nurses with AI?‘
The question goes to the heart of the debate. Nursing, like medicine and the wider health professions, is fundamentally relational. It depends on compassion, observation, communication, contextual judgement and moral responsibility. These are not optional additions to clinical knowledge; they are central to what it means to care for another person.
A chatbot may help produce a draft care plan, but it cannot assume professional accountability for that plan. The student or clinician must still ask the following:
- Is this information accurate and supported by evidence?
- Is it appropriate for this particular patient?
- What important information might be missing?
- Does the recommendation account for the patient’s preferences, culture, circumstances and values?
- Who is responsible if the advice causes harm?
These questions illustrate why evidence-based healthcare becomes more – not less – important in the presence of AI. Generative systems can communicate incorrect information with remarkable confidence. Future healthcare professionals will, therefore, need to distinguish fluency from truth and speed from sound judgement.
Communication matters just as much. AI can help a student learn something, but the professional still has to interpret health information, decide what to do with that information and then explain it to a patient or a colleague in a way that actually lands. That last step is where a lot of learning quietly happens, and it doesn’t happen in a chat window. It happens in front of a real person who is anxious, or hard of hearing or has already decided they don’t want the treatment you’re recommending.
Which is an argument for building in more face-to-face contact, not less. In the Hong Kong Healthy Longevity Study that we are currently conducting, first-year nursing students go into the community to conduct health checks with older adults. What they describe afterwards is rarely about the clinical measurements. It’s about learning to sit with someone, read a room and notice what isn’t being said, the kinds of things no chatbot can rehearse you through.
From AI adoption to professional responsibility
Responsibility for the use of AI cannot rest with students alone. More broadly, educators determine how learning activities are structured, and institutions establish policies concerning privacy, disclosure, assessment and academic integrity. Meanwhile, researchers evaluate whether these technologies meaningfully improve educational outcomes, rather than simply making tasks faster, and developers design systems that may influence how future professionals understand health, illness and patient care.
If these actors do not take up their responsibilities, poorly governed AI could reinforce bias, deepen inequities, weaken trust and encourage a dependence on automated answers. Responsibly governed AI, however, may broaden access to learning, support reflective practice and help students prepare for clinical environments in which algorithmic tools will be increasingly common.
In that context, the appropriate response to the rise of AI is neither resistance nor enthusiasm for its own sake. We should not reject AI simply because it is unfamiliar, but neither should we adopt it merely because it is powerful. Its educational value should be judged by whether it strengthens the qualities that healthcare professionals need: curiosity, compassion, critical thinking, integrity and accountability.
What we’re actually learning from all this
A few things have become clear now that AI has settled into nursing education, and none of them are especially surprising once they’re said out loud.
The first is that AI works best as a study partner, not a stand-in. It’s genuinely useful for understanding a concept you’ve read four times and still don’t get, or for checking whether your understanding of something holds up. But there’s no shortcut around the actual work: the reading, the thinking, the sitting with something uncomfortable until it makes sense. Students who use AI to skip that part aren’t saving time. They’re just deferring the moment they realise they don’t know the material in exams or practice.
The second is that fluent writing isn’t the same as correct information. This matters more in healthcare than almost anywhere else. AI is very good at sounding confident, which is exactly the problem. So, the habit worth building is one of slight suspicion: Is this actually accurate? Is it current? Does it hold up for this patient, in this context, with this set of circumstances? Accordingly, building up students’ and learners’ critical thinking capacity is more important now than ever.
The third takeaway is that students shouldn’t have to figure out the rules on their own. Right now, a lot of them are guessing, about what counts as cheating, what needs to be disclosed, whether it’s okay to paste patient details into a chatbot (it isn’t). That’s not really their failure. If responsible use is the expectation, it has to be taught: integrity, privacy, citation, bias and a frank conversation about what these tools genuinely can’t do.
And finally – the point that ties everything together – AI literacy isn’t a separate skill sitting off to one side. It belongs as part of the conversation already happening about evidence-based practice. The question was never, can students use AI? Most of them can, better than their lecturers. The question is whether they can use it in a way that makes their care safer, more ethical and more attentive to the person in front of them.
Which points to the bigger picture. The future of healthcare education shouldn’t be built around AI. It should stay built around people, grounded in evidence, held to ethical standards and honest about where the technology’s usefulness ends. AI can be a good learning partner, but it can’t take on the responsibility of thinking carefully, and it certainly can’t take on the responsibility of caring well. Those parts stay with us.
References
Gunawan, J., Aungsuroch, Y., Marzilli, C., N., Chaerani, E., & Montayre, J. (2024). Artificial intelligence chatbot as perceived by nursing students: A qualitative study. SAGE Open, 14. https://doi.org/10.1177/21582440241303453
Gunawan, J., Aungsuroch, Y., & Montayre, J. (2024). ChatGPT integration within nursing education and its implications for nursing students: A systematic review and text network analysis. Nurse Education Today, 141, 106323. https://doi.org/10.1016/j.nedt.2024.106323
To link to this article - DOI: https://doi.org/10.70253/ZYTA8958
Disclaimer
The views expressed in this World EBHC Day Blog, as well as any errors or omissions, are the sole responsibility of the author and do not represent the views of the World EBHC Day Steering Committee, Official Partners or Sponsors; nor does it imply endorsement by the aforementioned parties.
