An AI ate my Homework Assignment

A robot reading a magazine on a park bench

By Ted Murcray

A new artificial intelligence (AI) chatbot is taking the academic world by storm because this chatbot can create unique responses to homework assignments in a matter of seconds. ChatGPT is an open source (means free) online tool that was designed to be a chatting friend. However, it also has the capability of answering a host of questions and performing tasks. This skill allows the AI to create responses to paper prompts, and students are already using this feature to cut down on homework time to maximize play time!

https://chat.openai.com/chat

The work created by ChatGPT is unique. What that means if you can type in the same prompt multiple times and get different results. For example, I typed in this prompt, “In the style of Edgar Allan Poe, write a poem about climate change.” I received a well-written poem in about 5 seconds. A few minutes later, I repeated the prompt, and I quickly received another poem that was totally different from the first. This is important because SafeAssign and TurnItIn check for original work, not AI created work. Since each response is original (ChatGPT will not plagiarize itself), SafeAssign and TurnItIn will not flag the work. (There is also speculation that the developers of this AI used work submitted in systems checking for plagiarism to create the algorithm for this tool. I am not sure how accurate that is, but there is a growing voice arguing for faculty to stop using these systems to avoid future development of AIs. Honestly, it may be too late for that).

Weaknesses of Chat GPT

Several professors who have tried Chat GPT say the work created is typically C+ to B level work.  It is technically accurate and clean (will use proper APA style, etc.); however, the thinking provided in the work is not well connected nor original. 

Although Chat GPT can create in-text citations with appropriate reference lists, sometimes it is making up fake citations.  A professor would have to take the time to look up the references to see that they were not real, which is time-consuming, but he or she could easily see that the references were falsified.

Chat GPT does not have original thought.  It is creating original work by pulling from a variety of sources, but it is not creating truly original ideas.  One student tested Chat GPT’s ability to write code (which it can do) and solve coding issues.  He found that the AI was able to come up with the right process to solve a problem but often the solution was not complete or not completely accurate because it did not completely reflect the problem.  From this we know that the AI will have trouble with complicated issues.

What can I do as an instructor?

Our first step is to take a deep breath and put down any pitchforks you have picked up while reading this intro.  Some professors have started calling for a ban on all AI work at their universities.  This is not a productive path forward.  Instead, we must learn to work with this new tool.

Start by exploring Chat GPT yourself.  Go to the website and sign up for an account.  This may take some time as it is very popular right now.  I have found that logging in during the evening hours is the best time to get access.  Run any prompts you give students through Chat GPT and see what it comes up with.  Knowing what Chat GPT does with your assignments will help you plan.

Here are some first-step recommendations:

  1. Use a template for items like lesson plans, lab reports, etc.  There is currently no way to input a template into Chat GPT, so the AI will not be able to replicate that work.  Using templates that are found in your profession (especially if you are a professional program) will help your students acclimate to the profession and remove the possibility of using the AI to generate the product.
  2. Run your prompt through Chat GPT and provide it to your students.  There is no point in hiding this information from our students because they already know about it (I checked).  Instead, give it to your students to have them look at it.  Then, you can:
    1. Have your students analyze the work.  One professor who did this with her students said the students quickly realized the limitations of the paper.  It was ok, but not great.  They realized they could do better work on their own.
    1. Have your students use the paper you generated as the first draft.  Tell them they need to make it better by running track changes and adding comments to the paper.  They can then create a clean version, run it through Grammarly, and turn in a)the first draft with track changes and comments, b) the clean draft, and c) the Grammarly report.  This will allow you to see how the students considered the original work from the AI and worked to improve it with their own thoughts.
  3. Look for more workshops in the future.  The University Writing Center, led by Dr. Tae Sung, anticipated the release of a writing AI several years ago.  Since that time, the team has worked to shift ENG 113 and ENG 123 from being product-oriented to being rhetoric-oriented.  One of the key criticisms of Chat GPT is that it cannot handle rhetorical writing well.  The TLC is partnering with the UWC to create workshops for instructors to learn how to adjust our assignments to have a foundation in strong rhetoric, which will create better thinking from our students and reduce the usefulness of any writing AI.

Samples of work from Chat GPT

What you will likely note as you read through these examples (and others you may produce on your own) is that they are not terrific. You might also note that you have seen work that looks remarkably similar to this that was turned in to you. Note that in the Leadership Analysis example, I was able to tell the AI to include specific information from class discussions, and it was able to do so. Think about how this might affect assignments you have given to your students.

8 Comments.

  1. I agree with embracing the technology, incorporating it into classwork, and have the scholars engage in making it better and finding mistakes.

  2. I asked Chat-GPT how it could detect when it is being used, and it responded:
    “It is important to note that detecting AI-generated text can be challenging, and there is no foolproof method for doing so. However, by paying close attention to the quality and originality of the writing, and using plagiarism detection tools when appropriate, you may be able to identify AI-generated text.”

    These videos are interesting:

    Detecting Chat-GPT-3 generated text — https://youtu.be/XD1M3vavSAE
    AI -written peer-reviewed essay — https://youtu.be/Lto6exrpChQ
    AI personal research assistant — https://youtu.be/xeRJiCCpGro

  3. Thank you for the article so to anticipate the future of paper assignments. Fortunately, we don’t have many in PA. For the “template” items, do you suggest using a rubric so to have a breakdown of what is expected for an “A” quality work?

  4. Thank you for sharing this update. The recommendations of how to use AI within the classroom is a good idea. Looking forward to additional modules / continuing education on the topic.

  5. Thank you for writing this Ted, and for helping me find my way to it today. I am very interested in learning more and will try to watch for talks you and others are giving at CBU on this topic. Super interested too in learning how to change writing assignments away from product oriented to rhetoric oriented since that is something the device does not do.

Leave a Reply to Julie Cancel reply

Your email address will not be published. Required fields are marked *