AI in Higher Education: ChatGPT use cases

Millions of people have tried out ChatGPT, as I did last fall. After trying out different kinds of questions, and seeing both its strengths and weaknesses, some of you may have wondered how to put it to work for practical purposes. After taking a class on prompt engineering (more on that below), my ability to use ChatGPT grew significantly. Here are some of my use cases within the past few weeks, along with the prompt that I used to get started. In many cases, you will find that you have to do some trial and error in the prompt, or redirect or add a second prompt to get where you need to go. Note that I am using ChatGPT 4.0, which costs $20/mth, has more advanced logic capabilities, and you can upload and analyze files. Anything here that doesn't involve a file would probably be about the same in the free version.

Brainstorming on strategies and tactics to achieve an LSU goal. Prompt: You are an expert in higher education, specializing in student success. Suggest 10 strategies for increasing LSU's first-year retention.

Download and place all materials into a single file and then analyze the file. I wanted to find all the references in LSU's policies to a particular question. We have well more than 100 policies, all publicly available but not easily searchable. I wanted to put all of the policies together in a single file so that I could search all policies simultaneously. Prompt: Write a python 3 script that accomplishes the following tasks:

• Goes to a URL, where a long list of pdf documents are listed via web links;
• Downloads all of the pdfs;
• Concatenates all of the pdfs, in the same order as they appeared on the website, into a master pdf file.

After creating the script, which I ran using Anaconda/Python, it generated an error, which I fed back into ChatGPT and asked it to correct. About 15 minutes later, I had my single pdf file with all policies. In ChatGPT 4.0, I was able to upload it and ask it to summarize LSU policies on certain topics.

Summarize a long article. Prompt: Summarize the following article (followed by a paste of the article). Note that ChatGPT and other large language models have limits on the length of the article that can be pasted. There are ways around this limitation, but those involve writing some code and using OpenAI's API.

Summarize a long string of emails and tell me what my action items are (note: you should never upload anything confidential to ChatGPT - it is unclear if that information could be accidentally released by ChatGPT in another response). Prompt: I am going to cut/paste a long email chain into our conversation. You will summarize what each person's points are in the email chain. You will provide your summary as a series of sequential bullet points. At the end, list any open questions or action items directly addressed to me. My name is Roy Haggerty.

Write a draft syllabus for a class. Prompt. You are going to assist me in designing a 16-week course [describe class in 2 – 3 sentences]. Ask me questions that will help you design the course until I tell you to stop. You will ask me a question, wait for my response, and then ask me another question. Ask me the first question.

After answering a few questions (you can tell it to stop at any point), it will likely give you a class outline. At that point, you can ask it to turn the outline into a syllabus. In my trial with this, it gave me a syllabus that was about 50% ready-to-go... not a bad first draft. It still needed work, and all of the LSU-specific materials, but it got me going.

Analyze a data set. Prompt: You are an expert data analyst. I have a data set that [description of data set]. Suggest a statistical analysis of this to answer the question [question].

It will suggest a statistical analysis. At that point, use your judgment and do a bit of web searching (or ask back to ChatGPT) to understand more about the statistical method(s) described. You can follow that with a second prompt.

I am uploading the data set. Please do the statistical analysis you suggested, and provide your interpretation and summary of findings based on those analyses.

Prompt engineering. Prompt engineering is the design of a prompt, or a series of prompts, that assist the LLM to perform the task you wish to complete. There is a lot of good information on the web about this. I started by taking a short course offered by Dr. Jules White at Vanderbilt University offered via Coursera, called "Prompt Engineering for ChatGPT".

Ethics of using ChatGPT. As I've said before, in my opinion, using ChatGPT for routine matters such as analyzing a data set, writing an email based on points that you provide, summarizing some text, etc. no more require notification that you have used ChatGPT than does notification that you used excel or word. On the other end of the spectrum, where the task is creative (like the image at the top), or you are using ChatGPT to generate something new (a class syllabus), I think that a notification of how AI was used is appropriate. Also, as I've said in this article, never put confidential information into ChatGPT, because you can't be certain that the information will remain confidential.

Next week, I'll tell you how I think AI is going to change higher education. Stay tuned.