AI in assignments

While all professors and teachers must deal with Artificial Intelligence, I have never thought it a good idea to forbid students from using it. I also don’t think it’s a good idea to ignore it.

Whenever a new technological challenge to traditional teaching has arisen, the initial response (usually either enthusiasm or despair) has eventually quieted down to a norm. The question is how to accomplish this with AI.

First, we must be clear what AI actually is. It’s been in use for years, but I’ll just give one example: spell checking. At first, some were aghast at spell checking. How would students ever learn how to spell? A decade later, profs were insisting that all papers be spell-checked before being submitted. Grammar checkers weren’t far behind. Professors were tired of reading/correcting/mourning spelling problems, since their job was to focus on content.

Now the content itself is being created by computers — it’s generative AI. But instead of producing better content, it’s sometimes worse. Human error, as with all errors, are magnified. (Remember the old meme “to err is human — to really mess things up requires a computer”?) And it’s impossible to tell what a student is learning about the subject when they didn’t write the paper, so how do we grade work?

Why do we use AI?

At the most basic level, student use of AI for written assignments is plagiarism, with all the problems that entails. It is harder to tell whether a student has used AI, since unlike the old practice of lifting passages that were better written than a student could manage, the content comes out with errors and problems. We just don’t know who made them.

Students use AI for the same reasons they plagiarize generally: laziness, perceived lack of time, lack of confidence in their own understanding, lack of English skills. All of these are affective disorders, and all of them have causes often related to previous academic experience.

Of course, professors also plagiarize, or at the very least may rely on test banks, textbooks, and other resources not created by them. They do this for similar reasons: laziness, perceived lack of time, lack of confidence in their own understanding, and lack of English skills.

Changing the focus from product (the paper, assignment, or post) to process might help.

An example

I give an assignment asking students to use primary sources to construct their own historical argument in a short paper. I am aware that they may work with friends, buy a paper online, or use AI. The latter has become most common so the question is how to work with that process rather than against it.

(You’ll notice the assignment is designed to make students choose their topic. I don’t ask a question that can be typed into Google. And they have to use sources from the class, not just anything they find online. I don’t use “prompts”, but we practice the thesis/evidence thing a lot in class, and I model it.)

After crafting the assignment to minimize boilerplate responses, one could assume students will use AI. Instead of punishment, and spending hours trying to use AI to figure out whether the student has used AI, one could require documentation.

Instructions could go like this:

Write a three-page paper on what three of the assigned historical documents have in common, creating a historical thesis that’s proven with those documents. You are welcome to use AI at any stage, documenting your work by creating footnotes. These footnotes must contain which AI generator was used and list all search term/phrase/question efforts in order.

For example:

President Abraham Lincoln developed the Emancipation Proclamation in an effort to cause slave rebellion in the Confederacy.³

3. ChatGPT: Emancipation Proclamation, Emancipation Proclamation and slavery, Why did Lincoln write the Emancipation Proclamation?, Was Lincoln trying to cause slave rebellion with the Emancipation Proclamation?, Lincoln and Emancipation Proclamation, Emancipation Proclamation goals.

Skill sets

Those of us who mourn the loss of the card catalog and shelf browsing are aware that one cannot really browse anymore. Search is the new browse. Search terms are crucial; knowing which ones to use is an important skill for finding information. You can’t find the answers without asking the generative AI program the right questions.

Documenting process is meta. The student becomes more aware of how they’re searching. If we agree in class that AI is sometimes wrong, those who care may even fact check what AI tells them. Since what’s important to me is the skill in developing a thesis/evidence argument, I can then grade the product as I did before, on accuracy and writing and fulfilling the assignment (or whatever is on my rubric).

Is AI a horrible thing?

Sure it is. The magnification and popularization of inaccuracies, the pressure to work “smarter” and more superficially, the lack of deep thought and analysis — all of these are horrible.  When society values these things, a teacher can focus on ways of training the mind to counter them, rather than opposing them directly.

That’s natural intelligence in action.