Dr. Felicia Vulcu is an Associate Professor (Teaching Stream) in the Department of Biochemistry and Biomedical Sciences, Faculty of Health Sciences, at McMaster University. In her recent ‘Research Skills Laboratory and Inquiry’ course (BIOMEDDC 3C06), Dr. Vulcu has been utilizing ChatGPT to assist with the development of life-like scientific images to demonstrate what students can expect to see during their laboratory experiments.
In addition to these AI-generated scientific images, Dr. Vulcu is also introducing AI policies related student work for lab reports and has used ChatGPT to generate overviews of her class AI policies for students. See her article on this topic for additional information.
Dr. Vulcu has currently begun to experiment with the use of ChatGPT for the creation of scientific images for use in laboratory manuals, lectures, worksheets, and assessments (e.g test, quizzes, etc). Over the summer, she has been redesigning a range of teaching materials that will be introduced into her courses during the upcoming academic year. She wanted to explore if AI could generate life-like scientific illustrations that would provide students with greater clarity about scientific results and observations. She has been able to successfully create purpose-built images that align closely with course learning outcomes.
Some of her early applications have included:
Depicted below are two successful image generation efforts.
Firstly, the images on the left depict a gradual change in turbidity (cloudiness) of an E. coli cell culture over time. Having initially demonstrated this by manually creating a Microsoft PowerPoint image, Dr. Vulcu was able to replicate a more realistic looking series of samples with increasing cloudiness over time using ChatGPT. The goal was not simply to improve the appearance of the figure, but to provide students with a clearer visual representation of what they should expect to observe during the laboratory experiment.
Secondly, the images on the right depict an agarose gel electrophoresis result. Dr Vulcu originally generated the in silico experimental results and image using a software called SnapGene. However, she wanted students to be able to visualize the appearance of more realistic gel features such as slight lane smearing, subtle loading imperfections, variable band sharpness, etc. The successful AI generated version of the same experimental results presents a more realistic image that can help students build a stronger connection between the expected outcome and the appearance of actual experimental data. These AI-generated images are clearly identified as conceptual, fictitious teaching resources created for educational purposes only. These images are designed to help students interpret experimental results while also recognizing common features of real laboratory data. They will be incorporated into laboratory manuals, lectures, worksheets, quizzes, and assessments beginning in the upcoming academic year.
Although photographs of real laboratory experiments remain an important part of teaching, they can be thoughtfully supplemented with AI-generated images that help illustrate concepts, reinforce learning, and prepare students for what they may encounter in the laboratory. Some concepts require a carefully controlled series of images that can be time-consuming to capture. In other cases, such as agarose gel electrophoresis, students generate and interpret their own experimental data throughout the courses. However, real laboratory data cannot always provide every teaching scenario an instructor hopes to discuss. Generative AI has given Dr. Vulcu the flexibility to create purpose-built illustrations that showcase not only expected experimental outcomes, but also realistic troubleshooting scenarios and technical artifacts that students may encounter. Rather than replacing authentic laboratory experiences, these images are intended to complement them by helping students build confidence in interpreting scientific results before they encounter them at the bench.
While proving useful at generating life-like scientific images, ChatGPT does require several rounds of optimization prompts to produce the desired results which can still be time consuming (though less so than previous efforts). Dr. Vulcu is working to optimize her prompts to reduce the amount of time required to generate each image.
She has also begun to investigate the use of Vibe Coding (AI-assisted coding using written prompts) to develop interactive scientific simulations using Google Gemini Canvas. She has begun to develop an Agarose Gel Electrophoresis Simulator showing separation of DNA bands based on size.
Dr. Vulcu has found that AI can lower the barrier to producing an initial prototype, but the instructor still needs enough technical understanding to identify errors, revise the code, test the simulation, and ensure that it behaves as intended. She is developing her own knowledge in coding to enable her to create more accurate simulations. Her longer-term goal is to develop interactive activities in which students can manipulate experimental variables, observe simulated results, and explain the scientific relationships represented by the model.
—
Vulcu, F. (2026), AI-Generated Scientific Images for a Laboratory Course, Teaching in the Age of AI: Examples. Retrieved from AI-Generated Scientific Images for a Laboratory Course, Licensed under Creative Commons BY-NC-SA 4.0.
Teaching in the Age of AI
Teaching in the Age of AI, Updates