Cheating with AI and how to stop students from doing so is a major concern for educators.
But how widespread is it? Newly released data from a popular plagiarism detection company sheds some light on the issue.
And it may not be as bad as educators think.
Of over 200 million writing assignments reviewed by Turnitin's AI detection tools. Over the past year, the use of AI was detected in approximately 1 in 10 challenges, but only 3 in 100 challenges were primarily generated by AI.
These numbers haven't changed much from August 2023, when Turnitin released data for the first three months of use of the detection tool, said Annie Checchitelli, the company's chief product officer. .
“We reached a steady state and nothing has changed dramatically since then,” she said. “There are students who rely too much on AI. But it's not taking hold. This wasn't 'the sky is falling.' ”
The fact that the number of students using AI to complete their school work has not spiked over the past year is consistent with Stanford University's findings. It was released in December. Researchers there surveyed students at 40 different high schools and found that the percentage of students who admitted to cheating had remained flat since the advent of ChatGPT and other readily available generative AI tools. Did. In the years leading up to ChatGPT's release, 60 to 70 percent of students admitted to cheating, and that remained the same in the 2023 study, researchers said.
Turnitin's latest data release shows that 11% of assignments were run through AI detection tools, and at least 20% of each assignment had evidence of AI in the writing. At 3% of assignments, more than 80% of each assignment consisted of AI writing, which aligns well with what the company was seeing just three months after launching his AI detection tool. ..
Experts warn against clinging to cheating and plagiarism
But another survey of educators found that AI detection tools are becoming increasingly popular among teachers, a trend that worries some experts.
Questionnaire for junior high school and high school teachers According to a survey by the Center for Democracy and Technology, a nonprofit organization focused on technology policy and consumer rights, 68% have used an AI detection tool, a significant increase from the previous year. Teachers also reported in the same survey that their students are increasingly getting into trouble using AI to complete assignments. In the 2023-24 school year, 63% of teachers said they got in trouble when students were accused of using generative AI in school work, up from 48% the previous year.
Despite scant evidence that AI is accelerating the wave of cheating, half of teachers surveyed by the Center for Democratic Technology say they distrust students submitting original work because of generative AI. They report that their feelings have increased.
Some experts warn that fixating on plagiarism and fraud is the wrong focus.
Tara Natras, managing director of innovation and strategy at ISTE+ASCD, a nonprofit organization that provides content and professional development on educational technology, says this can get students into trouble when using AI tools. He said this has created an environment where students are afraid to talk to their teachers about their concerns. And the curriculum.
“We need to reframe the conversation and engage with students about how AI can support their learning, and how AI may negatively impact their learning.” she said in an email to Education Week. “Activities such as using AI to write an essay and pass it off as your own are detrimental to learning, while activities such as using AI to break down difficult topics to enhance understanding I want my students to know that activities help them learn.”
Shifting focus to teaching AI literacy and developing better policies
Students in the Stanford University survey said they generally believe AI should be used as an aid to understanding concepts, rather than as a flashy plagiarism tool.
Nattrass said schools should teach AI literacy At the same time, involve students in drafting clear AI guidelines.
Nattrass also recommends against using AI detection tools in schools. She said it was too unreliable to authenticate student efforts, and false positives could have devastating effects on individual students and create greater mistrust. Some studies have found that AI detection tools are particularly weak at identifying original writing by English learners. From AI-driven prose.
“Students are using AI and will continue to use it, with or without educator guidance,” Nattrass said. “Teaching students about the safe and ethical use of AI is part of our responsibility to help them become contributing digital citizens.”
AI detection software actually uses AI to do its job. These tools are trained on a large amount of machine and human-written texts so that the software can ideally recognize the difference between the two.
According to Turnitin's website, the company's AI detector claims to be 99% accurate in determining whether a document was written with AI, specifically ChatGPT, as long as the document consists of at least 20% AI writing. I am.
Ceckitelli pointed out that no detector or test, whether it's a fire alarm or a medical test, is 100 percent accurate.
While he said teachers should not rely solely on AI detectors to determine whether students are using AI to cheat, he argued that detection tools can provide teachers with valuable data.
“It's not conclusive evidence,” she said. “This is a signal that can be used in combination with other signals to start a conversation with students.”
As educators become more familiar with generative AI, the focus will shift from detection to transparency, Checchitelli said. It's about how students cite or communicate how they use AI. When should educators encourage students to use AI in their assignments, and do schools have clear policies regarding the use of AI? Also, what exactly constitutes plagiarism or fraud?
“The feedback I'm currently hearing from students is: 'I'm going to try it.' I'd like some more guidance on how and when to do it so I don't get into trouble. But,” Checchitelli said.
