Back when I was teaching 11 classes a semester as an adjunct (don’t judge), I created templates for giving students feedback. I taught these classes entirely online, so I tended to repeat myself. The templates had blank spaces for me to fill in the specific details relevant to an individual student’s topic, but gave me structure and transitions for responding in discussion boards and providing feedback when grading. 

Currently, I’m teaching a new-to-me pre-designed course (PDC), so I have a fresh set of assignments I’ll need to create templates for. I wanted to see if AI could help with this process.

In the past, brand new courses  would take me an entire round of teaching  before I could create my set of templates. Each template would have a variety of alternatives depending on the patterns I started to see in student responses. Normally, I’d change the assignment to address those issues, but in this case I wasn’t allowed to change the curriculum (in some cases, it was locked). 

This is a “do it with me” piece where I’ll walk you through how I used an AI writing tool to help me create a template for responding to my online students. I’m lucky enough to have access to the new AI writing tools in the PowerNotes platform, so that’s what I used, but, you can also, with some extra prompting, do this in ChatGPT. If you’d like to try out PowerNotes’ AI features for yourself, reach out here.

‍TL/DR: There is a point at which you begin to question how much time you’re spending trying to get the AI to give you what you want. There is also such a thing as too much information for the AI - it starts to get hung up on certain words/phrases and doesn’t consider the context of those words and phrases because it can’t. Knowing those two things can help you decide how much is too much and when to throw in the towel. 

‍

Getting Started: 

The first step I took was asking the AI to give me a template for discussion responses. I spent an inordinate amount of time prompting, and in the end didn’t use it for responses. I felt comfortable putting my instructions into the PowerNotes AI platform because the information I input is anonymous, held for 30 days before it’s deleted, and isn’t used to feed the ChatGPT beast.

My first prompt was:

Use the discussion prompt in the note, act as a college level information literacy instructor and, using a conversational tone, create a template that can be used to respond to individual students that have posted. Give me a template for students who have done all of the requirements and a second template for students who have not completed all the requirements. 

For the sake of space, I’m not going to include the discussion prompt here, but it asks students to select a topic from a predefined list and then respond to the following topics that are formulated as questions: 

Topic and keywords

1. Research question
2. Key concepts in the research question and alternative terms
3. Boolean logic phrase they might use
4. Any questions or things that need clarification

The first output was super long and read more like an email. It did give me two separate ones, but the first one was divided by each of the 5 questions and all of the language implied the student had done something incorrectly or it wasn’t complete. 

The second input is just straight up accusatory and focuses on not completing all of the requirements (and prompted me to include their entire discussion post), so clearly it wasn’t formatted as a discussion response or a template. 

I prompted it to:

Update the tone to come from the perspective of a compassionate teacher

The result was adding that I was a compassionate teacher.🙄 

‍

Starting Over

So I started from scratch with a new session and a new direction. This time, I asked the AI to:

Using the discussion board instructions in the note, act as a college level introduction to research instructor teaching an introduction to research course. Create a 150 word template, with bracketed notes with a prompt to add specific details, that can be used for instructor responses to an individual student’s initial discussion post response to those instructions. Connect the discussion instructions back to the course theme of learning to do research in the library, and ask one question to continue the conversation. Use a conversational tone.

I got similar results as the first round (including being way longer than 150 words). The biggest issue was the assumption that students weren’t completing the requirements, which is fascinating. Does this mean that the information the LLM (large language model) has suggests that feedback is given from a deficit model? Does this mean it can’t understand human learning (that it can occur even if things are “complete” and there’s no “deficit”)? 

So, I next added: 

Assume that students have followed the directions.

This shifted the blame from the student not doing it to the instructor forgetting. The table shows an excerpt from this second session from both the first and second prompts.

‍

Still not useful, so I got more specific: 

Assume they've included keywords, a research question, and boolean logic.

I continued to have the same problem with the output not being useful. For the third prompt, the AI gave me: 

Great job on selecting your topic and identifying key words during the scoping process. Could you please remind me of the topic you chose and share some of the key words you discovered? This will help us gauge the relevance of your keywords and ensure they align with your research focus.

‍

Starting Over Again

So I started a third session. In this session I decided to take out the full discussion prompt and summarize it myself: 

Act as a college level introduction to research instructor teaching an introduction to research course. Create a 150 word template, with bracketed notes with a prompt to add specific details, that can be used for instructor responses to an individual student’s initial discussion posts that ask students to list a research topic, keywords, a research question, some alternative keywords, and Boolean search terms. Assume students have completed all of the requirements in the list. Connect the discussion instructions back to the course theme of learning to do research in the library, and ask one question to continue the conversation. Use a conversational tone.

That, surprisingly, gave me something more useful. Not for the responses to discussion boards, but for feedback in the gradebook. The Brainstorm session gave me this: 

This wasn’t what I wanted for the discussion board, but it was a template that gave positive reinforcement and left space for me to personalize for students. I decided to stop trying to get it to give me something to start from to respond in the discussion board. 

For the discussion boards, I just focused on answering some of the responses to the 5th piece of the prompt: where do you still have questions or uncertainties. 

‍

Using the AI Output:

What I did find this output useful for was grading. So I took this template and tried it out as a feedback template on grades for this activity. The grading rubric was two items: 

1. For the initial post - the post is complete and thoroughly and thoughtfully answers the prompt. The options were meets, partially meets, or missing/not submitted.
2. For responses - responds to multiple classmates’ initial posts and those responses are thoughtful and contribute to the discussion constructively. The options were meets, partially meets, or missing/not submitted.

Not my favorite rubric, but not the worst, and it left me with room to use the template as the space where I focus on coaching students toward the next part of the class. 

I felt the need to first edit the template, though I didn’t change a whole lot because the AI had given me positive reinforcement statements and left space for me to insert things. It also largely assumed students had followed the directions and included everything. The following are my edits: 

Hi [Student's Name],

Thank you for sharing your research topic, keywords, research question, alternative keywords, and Boolean search terms. It's great to see your enthusiasm for exploring a subject in-depth. Your topic choice, [student's topic], seems of interest to you, which is an excellent place to start! 

Your keywords, [list of keywords], appear to be relevant and will likely yield valuable resources. Additionally, your research question, [student's research question], is clear and focused, which is essential for conducting effective research.

As for your alternative keywords, [student's alternative keywords], these can serve as excellent variations to expand your search and uncover diverse perspectives on your topic.

Using Boolean search terms, such as "AND," "OR," and "NOT," will allow you to refine your search and obtain more precise results. Make sure to combine your keywords strategically to enhance the relevance of your findings. You might try [insert personalized suggestion].

One question to consider moving forward: Have you thought about or found any specific databases or sources you plan to explore further?

Like with any template I’ve used in the past, as I come across a space where students need more personalized feedback or a very specific suggestion, I simply change the template as I’m responding to them. 

In some cases, if enough students are having the same issue, I’ll make a new template from the old. Normally, I’d just change the instructions, but when you’re teaching a PDC, you don’t often have that privilege. For example, a few students had closed questions, so that research question sentence became:

Your research question, [student's research question], is a good starting point, but remember your question should be open. This means that the answer should not be yes/no (or any other binary like good/bad, etc.). For your question, you could <I inserted specific suggestion>.

I also had a few students who needed more specific suggestions for Boolean terms, others who were spot on, and still others who went so far as to use asterisks, even though it wasn’t in the instructions. That means that sometimes I was changing the template from: 

Make sure to combine your keywords strategically to enhance the relevance of your findings. You might try [insert personalized suggestion].

to encouragement that they’d done well and a reminder that as they come across new potential alternative terms to not be afraid to pull them into the logic phrase. 

‍

What I Learned: 

I spent more time than I should have trying to make this work, but the resulting template for grading was worth my time. It would normally take me at least one run through of the entire class to develop the templates for each grade item. Now that I know pasting the whole prompt can make things go wonky, I can save myself time on the next round.

This experience also reaffirmed for me that we learn through failure.