Building Constructive Feedback Skills Using AI-Generated Content

Dr. Sarah Whitwell (she/her) is an Educational Developer in the Office of Community Engagement and a Sessional Instructor in the Faculty of Humanities. In HISTORY 4G03 / PEACJUST 4GG3: Nation and Genocide in the Modern World, Dr. Whitwell has been leveraging Generative AI to help students build their critical thinking and feedback skills. Each year, she creates a series of research essays using ChatGPT. Students are then asked to evaluate those essays, acting as teaching assistants to provide constructive feedback. After providing their feedback, students reflect on the value of Generative AI for conducting research on complex historical topics.

The Change

Recognizing that many of the students who take HISTORY 4G03 / PEACJUST 4GG3 are planning to attend graduate school, Dr. Whitwell endeavoured to create learning opportunities that would help students prepare for tasks they might encounter as teaching assistants: facilitating discussions, synthesizing complex arguments, and providing feedback.

Genocide is a topic that some students find offensive and / or traumatizing, so creating an atmosphere of mutual respect and sensitivity is key. Rather than having students critique the work of their peers, especially on highly nuanced topics such as what constitutes genocide, Dr. Whitwell decided to use Generative AI to generate passages for the students to evaluate. This way, students learn to provide constructive feedback, and they also gain experience working with and evaluating AI-generated content. Students are encouraged to draw their own conclusions about the usage of Generative AI for studying historical topics.

The New Process

Dr. Whitwell begins by generating a series of research questions related to course topics. For example:

• What is the relationship between genocide and gender?
• How has genocide been defined historically?
• How and why are certain genocides denied?
• How can truth, justice, and reconciliation be leveraged to move forward after genocide?

Dr. Whitwell then asks ChatGPT (free version) to answer those questions, adopting the persona of a high-achieving student pursing a postsecondary education. As part of the prompt, she also outlines expectations around word counts, primary and secondary sources, and citation styles. Basically, ChatGPT is given the same instructions that would be given to any undergraduate student when writing a research essay. ChatGPT then produces a short (~750 word) essay that students will be asked to evaluate.

By generating the research essays herself, Dr. Whitwell is able to ensure that the quality of the essays is relatively consistent across topics so students can freely pick topics that are of interest without worrying about advantages or disadvantages. Dr. Whitwell is also able to more quickly identify strengths and weaknesses in the passage ahead of time, which streamlines the grading process as students provide feedback on the passages.

For each topic, students are given both the prompt and output. The goal is to be transparent so that students can draw their own conclusions about Generative AI. It is important that students know that the instructor did not ask the AI tool to make any intentional mistakes or adopt a certain perspective.

Students are welcome to self-select which research essay they want to provide feedback on. Although intentionally broad, each research essay touches on key course themes that students have been exploring throughout the term. As such, students have the necessary background to evaluate essays dealing with these topics. Moreover, by the time the assignment is due, students have long been practicing how to identify the strengths and weaknesses of a research essay through weekly discussions on the course readings.

Once students have selected their passage, they provide both inline comments and summative feedback. Then, based on the feedback provided, they write a reflection on the value of Generative AI and what the exercise has taught them about writing a research essay.

Students are graded on both analysis and expression. Feedback on the passages should succinctly capture the overall strengths and areas of opportunity, balancing both positive and constructive comments. Students are also expected to provide feedback that is appropriate for the length of the paper, which helps them practice time management skills and learn that they might not necessarily comment on every awkward sentence. A-range papers are those that not only identify strengths and weaknesses, but articulate how to improve the overall passage.

For the reflection, students are evaluated on their ability to support their conclusions about the value of Generative AI with evidence. They are expected to make clear connections to the passage and the exercise. Students are graded on their ability to present an argument rooted in evidence, not an opinion based on feeling.

The Impact and Feedback

Dr. Whitwell has now implemented this assessment twice and received positive feedback from students in both iterations. Not only do students appreciate a break from the standard research proposal and research paper, but they also appreciate the opportunity to develop skills they will need as teaching assistants in graduate school. Both anecdotally and in the course evaluations, students expressed feeling more prepared to be teaching assistants after completing the assignment.

Another strength of the assignment is that it encourages students to develop their own informed opinions about Generative AI. The reality is that Generative AI is not going anywhere. Students will encounter the tool not only during their time at McMaster, but as they graduate and enter the workforce. They need to understand how Generative AI works, what it does well and not so well, and what it means to use Generative AI in ethical and sustainable ways.

Dr. Whitwell offers the following advice based on her experience:

1. Make sure that students understand how Generative AI works. Although students may be using Generative AI in various contexts, this doesn’t mean they understand how it works. If students understand how Generative AI is trained and generates responses based on a ‘next best fit’ approach, they can better identify strengths and weaknesses when analyzing outputs.
2. Do not show bias towards Generative AI. Every instructor has their own opinion about Generative AI. The purpose of this assignment, however, is for students to draw their own conclusions. If students can sense that their instructor opposes or favours Generative AI, they may feel pressured to align themselves with the opinion of their instructor in order to secure a good grade.
3. Expect to make changes. When Dr. Whitwell launched this assignment for the first time, she had a rough idea of what she was looking for in terms of grading. Did the students identify any issues with the citations? Did they analyze the quality of the thesis statement? Students certainly did these things, but to Dr. Whitwell’s surprise, many of the students were quite mean to their fictional students, criticizing them for their mistakes instead of offering constructive comments. While ChatGPT might not have feelings, the students they might grade in the future certainly will. So now she grades for tone as a component of the assignment, helping students understand how to actually give feedback that won’t demoralize the recipient and will encourage growth.

Ongoing Work

Because Generative AI is constantly evolving, Dr. Whitwell is committed to updating this assessment model for each iteration of her course. Typically, this means generating new passages. While prompts can be recycled, it is important to generate new outputs because Generative AI models are constantly evolving and being trained on new content. If students are going to develop informed opinions about Generative AI and its uses, they need to be working with up-to-date content.

It is also important to experiment with different Generative AI tools. Different tools – ChatGPT, Claude, Copilot, etc. – are designed for different purposes and trained on different data. Exposing students to these differences can help them better understand the new digital landscape.

One of the strengths of this assessment model is that it can be adapted for any discipline while still helping students develop the same critical thinking skills. While Dr. Whitwell piloted the assessment in a History / Global Peace & Social Justice course, it could easily be adapted to any discipline by adjusting the prompts.

—

Whitwell, S. (2026), Building Constructive Feedback Skills Using AI-Generated Content, Teaching in the Age of AI: Examples. Retrieved from Building Constructive Feedback Skills Using AI-Generated Content, Licensed under Creative Commons BY-NC-SA 4.0.