In 2023 a college professor based out of Texas made national news for failing his entire class. The offense? Suspected Usage of AI to complete classroom assignments.
The story made news not just because it was one of the first major potential incidents of artificial intelligence-related cheating, but because it was also incorrect. It turns out the professor had come to his conclusion by feeding student essays to ChatGPT, asking the AI if it had written the assignments itself.
Chat, GPT of course, said that it had because incorrect answers are a hobby of its.
The story, though both amusing and absurd, hints at a larger problem. Artificial intelligence is here. Students really can use it to cheat. Detecting this fraud actually is difficult. What does this mean for students and for teachers?
In this article, we take a look at the pros and cons of artificial intelligence in the classroom.
Industries that Will Suffer
In the sections below we will take a look at specific problems that AI poses in the classroom. For now, let’s think about how this technology could influence employment sectors of the future.
AI could eliminate the need for students to:
- Write
- Think critically
- Research
- Understand basic math, grammar, and other core academic concepts
Understand that we are not of the mind that any of the above skills are going away. Students simply now have a way to mimic them artificially. What job wouldn’t be affected by a roster of candidates who don’t know how to think on their own?
Nursing. Business. Education. These are all careers that depend on professionals who can think on their feet.
And what about the future of college education? Will students who got through high school with a lot of help from ChatGPT be prepared for college or even graduate school success? We simply don’t know—yet.
Below, we take a look at some specific problems that AI creates, and how teachers can work around them.
It’s Difficult to Detect
One of the major risk factors is that detecting AI use is very difficult. Yes, there are programs that you can use to run essays through. This software can say with some degree of accuracy whether or not AI was used to write an essay.
The issue is that a positive hit does not necessarily definitively prove that AI was used. It merely reflects that patterns common to generative AI were detected within the essay.
These patterns tend to be present in mediocre writing. Over-reliance on transitory phrases. Sentences without variance in length or content. Long, incoherent lists.
So while AI detectors can be used to identify suspicious behavior in students, it is difficult to issue consequences strictly based on a positive result.
A potential solution would be for teachers to identify criteria that they deem unacceptable in essays. Ideally, these guidelines would encourage content that would be difficult to produce through artificial intelligence.
The teacher could then use these criteria to judge essays. Not only will this make it harder for students to use artificial intelligence in class, but it also may produce stronger writers.
Naturally, determining what these criteria should be will be easier said than done. Still, educators have developed similar policies to control new technologies in the past.
The same way college professors had to create stricter rules around what sources were acceptable to cite in the internet era, teachers will now need to determine what an essay should look like when AI is able to generate hundreds of words a minute.
Clear citation guidelines could be effective in this capacity as well. While AI can generate citations, the student will need to double-check their work to avoid errors (artificial intelligence has been known to invent links when unable to find appropriate existing ones).
Other solutions could include:
- Requiring essays to pass AI detection.
- Including a presentation component to essays to ensure students are actually familiar with the material they submit.
- Heavy consequences that deter AI cheating.
With careful effort, educators can help ensure that their students complete work honestly.
Bad Research
Artificial intelligence can be used as a way to research topics more quickly. This is a more honest application of services like ChatGPT. Instead of spending hours in the library or trying your luck with search terms, you can simply ask a question and use its answer as a reference point for your essay. Sounds great, right?
Well, it would be great if generative AI was a reliable source of information. Anyone who has used ChatGPT or similar services extensively knows that this is not the case. Artificial intelligence applications have a reasonable degree of accuracy. However, they are not nearly dependable enough to build an essay upon.
For an AI-generated fact, to be worthy of inclusion essay, it would need to be confirmed by a secondary source, thus eliminating the time-saving factor of using AI in the first place.
Students who try to cut corners may wind up, making significantly more mistakes than they otherwise would have.
The citation guidelines that most universities use more or less nullify this concern. The majority of college professors require essays to come from pre-approved sources. ChatGPT certainly would not make that list. Grade school and high school students are not always held to high research standards.
For them, generative AI could have a negative influence on their research habits.
What Math Homework?
Admittedly, this is not a problem exclusive to generative AI. You can already cheat on math homework to a reasonably high extent just by using Google. AI just makes it a little bit easier than it already was.
Any student can now feed their math homework into a generative AI platform and receive not only the answer itself but any work that they would be required to show on the homework sheet.
Naturally, for a culture that is trying very hard to emphasize math and science in the classroom, this could be a considerable step backward.
It is also very difficult to control. Educators may get a good sense of which students are cheating with AI based on who has disproportionately high homework scores compared to their test results but even this is an imperfect science.
You can’t accuse a student of cheating just because they get good grades on their homework. What is the teacher to do? Unfortunately, there isn’t a good answer to that yet.
Are There Any Positive Applications of AI in the Classroom?
Sure. Adaptive learning is not a bad thing in its own right. Taken in moderation it can help students learn in a way that is most natural to them. Modern students already spend a significant amount of time in front of screens.
Often they will learn through interactive games and applications. One teacher can’t create different lesson plans to accommodate the needs of every one of their students. An AI platform potentially could.
AI could even be used to a productive end in the context of research. We highlighted earlier what bad AI-generated research looks like. But what about the college student who uses AI to generate a list of journal articles related to the topic they are writing about?
Instead of combing, endlessly through academic journals, they could have a list of high-quality, peer-reviewed sources in a matter of seconds. They would still need to read the articles themselves, but they may accelerate the research process.
With most things relating to AI moderation is key.
It’s been said that you can think of artificial intelligence as a reasonably intelligent assistant. It does make mistakes. It can sometimes be unreliable. However, it is affordable. Fast. Generally accurate.
When used sensibly and with a degree of caution, AI can accelerate education workflows, and help people learn in a way that makes the most sense to them.
Is that how it will be used? Time will tell.