Caught Between Innovation and Integrity: Generative AI Reshapes Student Writing

Since the emergence of generative AI, students across the globe have been misusing the technology to covertly cheat on their assignments.

According to a survey conducted by Brian W. Stone, 41% of students admit to utilizing AI in ways explicitly banned.

In the past, teachers could run a student’s paper through plagiarism-detection software that scanned the internet for identical or substantially similar text. 

There was no way for me to outright ban AI in the classroom . . . Students were going to use it

Unlike traditional plagiarism, generative AI draws on patterns learned from large datasets to produce new text that may not directly match existing online sources. Because the text is technically original, plagiarism-detection software may not identify it, allowing students to present it as their own work. 

Nick Bennett, English professor at the College of Lake County (CLC) said this makes enforcement of anti-AI policies difficult for educators. 

“There was no way for me to outright ban AI in the classroom,” Bennett said. “Students were going to use it outside the classroom if they wanted to.”

So far, the primary method used to catch students dishonestly using generative AI is AI detection software. Programs such as GPTZero or TurnItIn scan a passage for predictability and sentence structure variation to estimate how much AI content it believes the passage contains.

Despite their growing use, these programs are far from foolproof. 

“AI detection software is a little iffy, and often comes up with false positives, so students will be accused,” Bennett said. “They’ll say that they didn’t do it and then the school is kind of left unsure what to do.” 

Jay Castillo, a current students at CLC, said he has personally run into problems with these detectors. 

“This literally happened to me last night, I used Grammarly and got an AI detection for it afterwards,” Castillo said. 

Castillo believes AI detectors are useful as a preliminary check, but that teacher’s should still be responsible for going over the work themselves. 

“Students are very hesitant . . . they’re getting mixed messages between the different classes.

AI technology is still experimental, and many students who are genuinely trying their best, are unaware of the complexities that come with these tools.

CLC student Nina Sax said some students simply do not understand how different AI programs work, which can lead to a positive detection score on software like Turnitin.

Sax said she was accused of cheating after running her paper through AI for her English class. 

“When it was very new, I didn’t know it took my work and rewrote it for me. I thought it was just rephrasing. I thought it was just correcting my grammar.”

She later had to explain to her teacher that she was unaware she was being academically dishonest. Although she was allowed to redo the assignment, other instructors may not be as forgiving. 

Bennet said this confusion makes it difficult for students to know when or how they can utilize AI tools for academic purposes. 

“Students are very hesitant because they feel like it’s a form of cheating, even if a teacher says it’s okay,” Bennet said. “I also feel like they’re getting mixed messages between the different classes.

Many students have also deliberately generated complete works with AI, and some have successfully passed the work off as their own. 

Castillo said he was aware of a student who got a scholarship by using AI to write an essay.

“I talk about it as a tool, mainly for revision, for getting suggestions for making changes . . .

For the classes he teaches, Bennett’s stance on AI is not entirely negative.

“I talk about it as a tool, mainly for revision, for getting suggestions for making changes,” Bennet said. “I recommend things like Grammarly, that are less generative AI and more about sort of like editing software.”

Bennett’s said he also requires students to do the skeleton of their assignments in class. So, much of the work will already be done by the time they tie it together at home.

“I moved a lot of the writing process into the classroom with daily in-class writing assignments so I can help students get their piece off the ground,” Bennett said. “But when it comes to putting the essay together and then revising it, that’s still got to happen outside the classroom.”

Every professor has their own stance on AI, deciding for themselves how it may be used in the classroom, if at all. 

“I’m doing what I can to adapt to it,” Bennett said. “I’m not outlawing it. I’m asking that they let me know when and how they’re using it as a tool and I’m showing them how it can be useful as an editing technique… I don’t know if I’m doing it right. I don’t know if there’s a right way to do it.” 

“. . . the skills of putting together an idea and expressing it effectively . . . those are the skills that you start to lose if you lean heavily on generative AI.”

While Bennett is open to students using AI as a tool, he worries that overreliance on it can undermine the development of essential writing skills. 

“The building block assignments where you’re building the paper up, that’s where you’re learning the skills of putting together an idea and expressing it effectively, and those are the skills that you start to lose if you lean heavily on generative AI.”