Dr. Sara Andres is an Associate Professor (Research Track) in the Biochemistry & Biomedical Sciences program in the Faculty of Health Sciences. She teaches the Biochemistry and Biomedical Sciences: Structural Determination and Analysis of Macromolecules (3Z03) course. Historically one of the course assessments required students to write a Structural Analysis Report on protein structures and their related functions. As a result of Generative AI being capable of writing content to this effect, Dr. Andres decided to purposefully integrate AI into the existing assessment.
Dr. Andres goal for the AI-integrated assessment was to teach her 100 student class how to critically evaluate AI output while retaining the student’s ability to decide what content was relevant and important for the analysis. She wanted them to learn how to use AI as a tool rather than a replacement to their work. As such, she decided to have students employ an AI Large Language Model of their choice (e.g ChatGPT, Claude, Gemini, etc) as an editor of their initial work.
In this early experiment with AI-integrated assessments, students were first tasked to complete a portion of AI-free work. They had to to choose a protein structure that interested them from the scientific literature, search the Protein Data Bank for it’s 3D structure, and then examine it to understand how the structure contributes to the function of the protein. Using this structure and other referenced scientific literature they were required to write a first draft of their Structure Analysis Report. In the report they had to answer questions like:
They were also required to create a human-generated figure (using Chimera or PyMol) that demonstrated the 3D protein structure and key residues they described in the report.
After completing their human-generated draft, students were then provided with the following instructions to use their chosen Generative AI to help edit their work.
5. Submit the rough draft of the written report to and decide on the prompt to use to have ChatGPT edit your report for quality of writing – clarity, organization, sufficient detail. The prompt should not include any instructions related to adding scientific information. 6. Submit your prompt to ChatGPT, the initial draft of your report (likely part of your ChatGPT submission), and the output provided by ChatGPT. 7. Edit what ChatGPT suggested, making sure to correct any scientific inaccuracies, adding references, and additional details if desired. Save a copy of the report with “track changes” enabled to show what was changed between the ChatGPT output and your final submission, or mark-up a copy of the ChatGPT output to show what you changed. A minimum of 20% of the ChatGPT output and your final submission should be different.
5. Submit the rough draft of the written report to and decide on the prompt to use to have ChatGPT edit your report for quality of writing – clarity, organization, sufficient detail. The prompt should not include any instructions related to adding scientific information.
6. Submit your prompt to ChatGPT, the initial draft of your report (likely part of your
ChatGPT submission), and the output provided by ChatGPT.
7. Edit what ChatGPT suggested, making sure to correct any scientific inaccuracies, adding references, and additional details if desired. Save a copy of the report with “track changes” enabled to show what was changed between the ChatGPT output and your final submission, or mark-up a copy of the ChatGPT output to show what you changed. A minimum of 20% of the ChatGPT output and your final submission should be different.
Importantly, the student’s prompt was not allowed to request for the AI to add any scientific content to the article, ensuring that the student was required to do their own scientific research for their first draft. Furthermore, students were required to submit their full conversation with the Generative AI tool, enabling TAs to check that this instruction was followed.
After obtaining the Generative AI recommendations for editing the content (often in the form of an AI-generated re-write), students were required to make decisions about the validity, usefulness, and accuracy of the edits. By having students compile a final edit of the AI-generated content using the Track Changes function in MS Word, TAs were able to determine which edits the student accepted and which edits they decided to correct.
Students and TAs were provided with a rubric for the assessment, in which marks were divided between the scientific content of the analysis (10/35 marks), the human-generated figure (10/35 marks), the quality of the writing (5/35 marks), and the final editing of the AI-generated content (10/35 marks). The rubric for this final section can be seen below.
Dr. Andres was able to implement the new activity and assessment without any significant challenges.
To some students’ surprise, the initial prompt they provided the Generative AI with to edit their content resulted in the AI removing significant amounts of their initial writing. By requiring the students to re-edit the output of the Generative AI tool, this encouraged them to think critically about what content was genuinely important for their analysis. For the students who had significant cuts in their report, they had to critically decide if the AI had only cut material that was unnecessary for the clarity of their writing, or whether the content had still been an important part of the analysis.
Interestingly, the activity did provide an indication of which students were most competent with the use of the AI tool as a supportive editor. A small proportion of the students demonstrated the ability to repeatedly prompt the AI to make suggestions for editing, instead of writing the article for them, enabling them to create higher quality outputs with their own unique voice and writing style. These students were not required to edit their final report by 20%, as they did not ask AI to generate the actual writing, but instead wrote the article themselves, and only asked the AI for feedback and suggestions on what to improve.
Dr. Andres found the AI editing assessment to be a valuable tool in demonstrating to students some of the opportunities that generative AI provides, but also some of their limitations.
If she repeated this assessment in the future, she would repeat the same protocol but with some modification. Instead of having students prompt the AI to not add any additional scientific information, she would instead omit this instruction and require students to assess any new information the AI added for scientific accuracy. This would further increase their critical thinking around the important of accurate scientific data.
With the rapidly increasing capabilities of generative AI tools, Dr. Andres would also run the assessment with the roles of the student and the AI switched. She would have the students create initial prompts for the AI tool to write the initial report and then require the students to fact-check and edit the report to their satisfaction. She feels that this is representative of the methods students currently use their AI tools for support with their work.
Finally, Dr. Andres would encourage students to experiment with the use of generative AI for the generation of their scientific images and figures. She would require them to generate their report figures using an AI tool, analyze the figures for scientific accuracy, and then demonstrate the ability to prompt the AI to fix any flaws. By doing this, she would encourage students to use critical thinking and decision making for both AI text generation and image generation.
—
Andres, S. (2026), Employing Generative AI as an Editor for a Protein Structure Analysis Report, Teaching in the Age of AI: Examples, Retrieved from Employing Generative AI as an Editor for a Protein Structure Analysis Report, Licensed under Creative Commons BY-NC-SA 4.0.
Teaching in the Age of AI