Artificial Intelligence Image Generation as a Teaching Strategy in Nursing Higher Education

Artificial Intelligence (AI) is among the most influential technological advances of contemporary times. It stands to dramatically change the landscape of modern life, including online learning in higher education. Truly, the effects can already be felt in this very class. As educators and researchers, we must investigate the effects that AI has on our educational endeavors and stay informed as to the correct and ethical usage of such technology (Reed et al., 2022). Among our objectives should be developing usage for AI that maintains the integrity of student critical thinking. Rather than refuse AI, we must devise integrations that do not dilute the curriculum or delegate student work to the machine.

The nursing industry is not unaffected by the emergence of AI. It is already present in electronic health record review, physician SOAP note creation, medical chatbots for patients, and virtual nursing assistants for managers (Reed et al., 2022). A study by Reed et al. (2022) examined the use of image-generating AI, a form of generative AI, to create new images or media to enrich the nursing student experience. Many academic institutions have banned generative AI – most notably the use of ChatGPT to write papers – but many more are investigating how to embrace it (Reed et al., 2022). There is no doubt that at the least, understanding of the technology should be achieved by educators, so that their decision to include or exclude it are founded.

The study by Reed et al. (2022) was an IRB approved qualitative case study design at a public Midwestern nursing school. The participants were undergraduate nursing students in their final year. The students were given the following prompts to respond to in writing, then to use their responses to generate AI-created images.

  1. Describe the nursing profession.
  2. Describe why you want to be a nurse.
  3. Describe your biggest fears about becoming a nurse.

After the images were created, the students reflected on their experience. In general, review of the data found that students had a generally positive emotional response to the generated images, felt as though self-reflection on moral and professional identity was encouraged, but did think that the images were “strange” or biased (Reed et al., 2022). Overall, it was an interesting use of the technology. While not very scientific, it promoted arts-based pedagogy and adoption of technology.

Finding this article was especially interesting to me, because I used ChatGPT to assist in the creation of my didactic teaching plan in our previous assignment. Namely, I used it to help generate case study characteristics of patients undergoing hemodynamic instability due to acute-on-chronic heart failure and sepsis. I also tried to generate a coloring book page, but was unsuccessful, as evidenced by the dialogue below:

The prompt I gave ChatGPT 4.0 Dall-E-2: “Generate a coloring book page that teaches the path of blood through the hearts”. Resulted in the following image and text:

“I’ve generated a coloring book page that illustrates the path of blood through the human heart, designed for both educational clarity and engagement. This page includes labeled parts of the heart such as the atria, ventricles, valves, and major blood vessels, with directional arrows to show the flow of blood. It’s suitable for a wide audience interested in learning about the heart’s anatomy and function. Feel free to use it for your educational purposes.” (OpenAI, 2024)

As can be seen in the image, AI got the physiology of the heart very, very wrong. It seemed like it was trying to emulate an illustration from Gray’s Anatomy, but the important parts (the path of blood) was not correct at all. So I decided to try a more formulaic and less visual prompt:

“Generate a concept map outlining the 1-hour sepsis protocol as dictated by the SCCM.” ChatGPT responded as follows:

“I’ve created a concept map for the 1-hour sepsis protocol as outlined by the Society of Critical Care Medicine (SCCM). This map visually organizes the steps involved in the protocol, from the identification of sepsis to the initial resuscitation and management steps, including key actions such as measuring lactate levels, obtaining blood cultures before antibiotic administration, administering broad-spectrum antibiotics, and initiating fluid resuscitation. This map is designed to be clear and concise, suitable for healthcare professionals looking to quickly understand and apply the protocol.” (OpenAI, 2024).

Again, it can be seen that the image that was generated by the AI was very convoluted, with misspellings and no clear path – and a labels in Latin…? However, the response text more or less was correct in its content, but not something a clinician would ever want to rely on.

The research article and my experimenting with ChatGPT brought to mind an excellent way of integrating AI into an educational plan while not allowing the critical thinking content to be AI-generated. Basically, students would need to generate an original image like I did, but then research and relay the ways that the image is incorrect. This would include engagement of their faculties across Bloom’s taxonomy, integrating AI without letting AI do the actual work of the assignment. I think this could be a very successful integration of AI integration into higher education.


OpenAI. (2024). ChatGPT (Version 4) [Large language model].

Reed, J., Alterio, B., Coblenz, H., O’lear, T., Metz, T. (2023). AI image-generation as a teaching strategy in nursing education. Journal of Interactive Learning Research 34(2). 369-399. Retrieved February 28, 2024, from

Share This!

Scan each other's personal QR codes to earn additional points.