Using Chat GPT to Help Write Assessment Feedback
Walkthrough Contents
Introduction and templates
This is a walkthrough of how I use Chat GPT to support writing summative assessment feedback for the students in my institution. I have referred to templates I have built in this walkthrough, all of which are placed below. A key principle for me is not dehumanising ourselves as professional assessors in this context, therefore as you will see in the next heading, I am using this to save time, standardise feedback comments and feedback structure, and to identify feedforward points quickly.
The problems I am trying to fix
I am trying to fix three issues that has been apparent in my feedback practices over the previous several years. These are:
Reducing the amount of time it takes to write summative feedback comments for students.
Removing / reducing the repetitive nature of some feedback which I provide to students.
Standardising feedback structure and feedback language, particularly as we typically have multiple assessors per assessment.
The feedback comments are still generated from my judgements and my thinking.
Creating assessment rubrics
I have seen a number of different ways to present marking criteria to students, from generic descriptors through to purposeful modelled examples with clear explanations for students. To help me with using Chat GPT, I have landed on creating assessment rubrics for each assessment. These are still generic, but are slightly more personalised for the module content. This allows me to divide each criteria into different sections, and provide feedback statements for each grading boundary for each criteria. Below is an example of an assessment brief and a personalised assessment rubric I use.
Assessment brief and marking rubric example
If you are unfamiliar with this and it seems like a lot of work, you can get Chat GPT to help you create a statement for each grading boundary. What I do from this is simply copy and paste across the grade statement into the AI Student Feedback Template document. For example, if I judge the student to be 60-69% for their knowledge and understanding, I will copy and paste across that specific statement into the template.
Using prompts in Chat GPT
Using prompts in Chat GPT is crucial to generate good responses. My line of thinking with this is "Put good stuff in, you get good stuff back. Put bad stuff in, you'll get bad stuff back." Spending time creating good prompts is a worthwhile use of our time.
In relation to assessment and feedback, what I recommend with the prompt is:
Include the programme title as Chat GPT will use this to inform its response.
Include the level of the module. Levelled feedback is important for students to help them understand and take forward feedback comments appropriately.
Include the module title as Chat GPT will use this to inform its response.
Include the feedback structure you want it to produce. For example, any headings, any particular tasks.
Do not shortcut the prompt stage. Below is an example from my teaching:
You are an undergraduate lecturer leading a level 6 Inclusive Coaching in Sport and Physical Activity module on a BA (hons) Sport Development and Coaching programme. Students have submitted an essay where they were tasked with critically evaluating effective coaching strategies within a disability sport setting. You are tasked with providing feedback to students. Your feedback will be structured in the below format:
Essay feedback
Knowledge and understanding
Intellectual skills
Practical skills
Transferable skills
You will be provided with comments from the rubric criteria which align with their grade awarded for each of the four categories above, in addition to specific comments. Your task is to turn this into feedback and feedforward comments for students for each of the four categories. Your feedback must use the four headings above. Your feedback must also begin with three concise action plan points for the student to focus on to improve their knowledge and/or essay writing skills.
This prompt isn't perfect as I have to correct Chat GPT over the next few messages until I get to my desired feedback structure, however this has become my go to prompt as it gets Chat GPT close to my vision.
I have created the above template. You will need to change the prompt, and then you can copy and paste the template as many times as you'd like to. You will only need to change the heading to the student name, and by opening the navigation pane in Word, you will be able to quickly click through students. All I then do is copy and paste generic comments across from the rubric I create (see below), and add any specific comments that I think are important for that criteria piece. For example, if I judge that their knowledge was good in a specific area, I will add that in note format.
Chat GPT creating feedback and feedforward
After copy and pasting across all rubric statements and adding some specific comments for each criteria, I then copy and paste this into Chat GPT. If the prompt works well, you will get the feedback text you desire. If it doesn't provide you with the structure or comments that you want, you can then engage in dialogue with Chat GPT to get it to tweak its output. Having patience is important here. Once I get the first one right, I can copy and paste this into Turnitin and repeat the process for all other students. The good news is, once you get the first one correct, you can remove the prompt going forwards and replace it with a line:
Perfect, now use this feedback structure for all future students. Do the same for this students feedback.
Creating themes from feedback
A key benefit of using Chat GPT is saving time, however by the text being within a single chat you can generate themes of student strengths and common pitfalls easily. For example, at the end of all feedback being added and copy and pasted out to Turnitin, I will then add the following prompt:
Use all of your feedback above to provide themes for what students did well and what they could improve on. Identify five strengths and five areas for improvement
This allows me to take those themes and show current and future students, as shown below.
Summarising how I use Chat GPT for student feedback and feedforward
As a summary, these are the benefits and challenges I'm seeing from using Chat GPT for student feedback and feedforward:
My marking time has reduced by approximately 50%.
My feedback structure is now consistent, including when there are multiple markers.
It's easy to identify themes, which is useful for students, but also informs my teaching.
Lots of copy and pasting - I'm sure this will reduce with AI advancements soon.
Time needs to be invested into creating good assessment rubrics and templates early.
External Examiner and student feedback is positive.
Overall, I am reducing my admin time, increasing student feedback quality, improving my teaching and learning impact, and having more time to add even more value!
Chat GPT can be scary. Why not have a go. Reach out if you'd like to chat about it more!