This article was updated in June 2024.
Last summer it finally happened to me: While reading a student essay in an introductory online college course at one of the universities at which I teach, I began to suspect not only that this student hadn’t written it but that no human mind had. In other words, it was the work of AI (artificial intelligence).
Like many educators across the globe, I was thrust into a brave new world of modern teaching that I had not been trained for and to which an appropriate response was unclear.
While I was aware of many effective AI detection tools , I had also heard horror stories of false positives from these same tools. My institution had recently announced a policy forbidding instructors from accusing students of using AI in their work. This gave me some institutional framework in which to operate, but I also did not want to reward a student with a positive grade for using a machine to do their writing for them.
I’ve since encountered similar situations many times over the past nine months. During that time my AI radar has improved, and my understanding of the thorny questions around AI submissions has evolved. Here’s how I handled it initially and how I’m handling these cases today. Keep in mind that this is an ever-evolving process and appropriate responses are often institution, class, and student-specific.
1. Detecting AI: I Trust My Instincts to A Point
The first step in this process for me was to use common sense while reading student work. AI tools have gotten good at writing in one sense but tend to be bland – the written equivalent of white noise or smooth jazz. In addition, the student I first suspected submitted work that was markedly different from any they had submitted previously that term.
In the months since, I’ve noticed a trend of my online students never submitting their own writing, so this change in tone and style is often unavailable as a comparison point. On the flip side, I’ve become aware of a number of AI “tells” and have developed methods for detecting AI writing without technology. I’ve started using these more and more instead of AI writing detectors, which I’ve increasingly found to have too many false negatives, as there are now AI tools designed to read AI written content and tweak it to throw off AI detectors.
But as good as I think my spidey sense for detecting AI writing is, it would definitely not stand up in a court of law, so I do not make AI accusations based on this. Instead, I use these suspicions as a conversation starter.
2. I Confer With Colleagues
I have written about a student who was falsely accused of using AI to generate papers, so I know I need to proceed cautiously to be fair to my students.
I typically reach out to a colleague in my department who oversees my work to share my concerns and any questionable work. If it still looks suspicious, we then bring in a superior for a third opinion and to discuss a course of action. This process helps protect the student by ensuring that I’m not misreading the situation somehow.
Talking with others also protects me as an untenured educator because it documents that I am acting in accordance with the university’s AI’s policies. This is especially important because at many institutions these policies are evolving and in some cases haven’t been formalized yet.
Conferring with colleagues remains one of the most important steps with each and every AI paper I encounter. It can be time-consuming for me and them, Al but I think it’s important, as my sense of what is and isn’t AI writing is not perfect and students deserve the benefit of having this process be conducted with some oversight.
3. I Try to Speak With The Student Without Accusing Them of Using AI
My colleagues and I typically decide the best course of action is to try and speak with the student. During any conversation I ask how they conducted their research for the paper without mentioning AI or accusing them of anything.
I usually reach out via email and ask to talk with the student on the phone or via video call. If a student doesn’t respond, I email specific questions about the content of the paper, and say that I need these to be answered before I can grade their work.
At the same time, I also send class-wide emails that I had encountered suspected AI-use in the classroom. These announcements remind students that using AI writing for essays is against course policies.
Many times, the student will not respond, and I give the paper a zero.
I’ve encountered several instances of suspected AI use and most play out pretty much the same way. None of my students have been penalized for violating academic standards but none receive credit for work that they did not produce, which is my primary concern.
It’s a small, but I think important, victory for human writing.