The "ChatGPT" challenge faced by universities

Last semester, philosophy professor Darren Hick discovered another incident of cheating in his classroom at Furman University, and he posted an update to his followers on social media: " Alas, I caught my second plagiarist using ChatGPT."

Friends and colleagues responded, some with wide-eyed emojis. Others also expressed surprise.

"Only two?! I've caught dozens," said Timothy Mayne, a writing professor at Conestoga College in Canada. "We are in a state of full-blown crisis."

Almost overnight, ChatGPT and other AI chatbots have become the go-to source of college cheating.

Today, educators are rethinking how to teach a variety of courses in the fall, from Writing 101 to computer science. Educators say they hope to take full advantage of the new possibilities this technology brings to teaching and learning, but when it comes to assessing students, they believe it is necessary to design ChatGPT's "difficult" test questions and assignments.

For some teachers, that means returning to paper-and-pencil testing after years of purely digital testing. Some professors will ask students to submit edited version histories and first drafts of their creative process to demonstrate that their assignments were written with thoughtful thought and genuine effort. Other teachers were less concerned, saying some students were always finding ways to cheat and this was just the latest option.

The proliferation of AI-generated chatbots, including ChatGPT launched last November, has raised new questions as academics work to ensure students not only get the right answers but also understand how to get the job done. Educators say there is consensus on at least some of the most pressing challenges.

"Are AI detectors reliable?" Stephanie Laggini Fiore, vice provost at Temple University, said not yet. This summer, Fiore was part of a Temple University team that tested the detector used by Turnitin, a popular plagiarism-detection service, and found its accuracy to be “quite deficient.” She said it performed best at recognizing human assignments, but was spotty at recognizing chatbot-generated text and was least reliable on mixed assignments.

Could Students Be Falsely Accused of Cheating Using an AI Platform? Absolutely. In one case last semester, a Texas A&M University professor falsely accused an entire class of using ChatGPT for a final assignment. Later, most of the class members were proven innocent.

So, how can educators determine if students are dishonestly using AI? It's barely discernible unless the student confesses, as Schick's two students did. Traditional plagiarized texts are often similar or identical to the plagiarized original text, however, AI-generated texts are unique every time.

Sometimes the cheating is obvious, and Mayne, the writing professor, said he has encountered students turning in assignments that were clearly copied from elsewhere. "I've had a response that was: 'I'm just an AI language model, I'm fine with that,'" he said.

Last semester, in his required freshman writing class, Mayne logged 57 instances of academic honesty, a sharp increase in academic dishonesty from about eight cases each of the previous two semesters. AI cheating accounts for about half of that.

This fall, Mayne and her colleagues are overhauling the school's required freshman writing curriculum. Writing assignments will be more individualized, with students encouraged to write about their own experiences, opinions and perceptions. All assignments and syllabi will have strict rules prohibiting the use of artificial intelligence.

University administrators have long encouraged faculty to explicitly set ground rules.

Many schools are leaving the decision on whether to use AI in the classroom up to teachers, said Hiroano Okahana, director of the Future of Education Lab at the American Council of Education.

At MSU, faculty members can teach as they see fit Outlines are selected and modified. Help develop new mandates and policies.

Hart-Davidson said: "A question like asking students, 'Tell me in three sentences what is a Krebs cycle in chemistry?' will no longer be effective because ChatGPT will give a perfectly correct answer.” He suggested asking the question in a different way. For example, give a description with an error and ask students to identify the error.

There is growing evidence that chatbots have changed the way students learn and seek information.

Chegg Inc., an online platform for homework assistance, has been mentioned in multiple AI cheating cases. Chegg CEO Dan Rosensweig said the company's shares fell nearly 50% in the first quarter of 2023 due to a surge in student use of ChatGPT. Students who would normally pay for Chegg's services will now be able to use the AI platform for free, he said.

This spring, Temple University Librarian Joe Lucia said that with the advent of chatbots, the use of research tools such as library databases has declined significantly.

"Students seem to see this as a way to quickly find information without expending energy or time using dedicated resources, saving a lot of the process that would be required," he said.

Shortcuts like this are worrisome in part because AI can easily make things up, and this “fiction” can lead to “distortion” of information. AI developers say they are working to make their platforms more reliable, but it's unclear when or if that will happen. Educators also worry about what to lose if students skip learning steps.

"There's going to be a big return to paper-and-pencil testing," says Bonnie MacKellar, a professor of computer science at St. John's University in New York City. She said the subject already had a "massive plagiarism problem", with students borrowing computer code from friends or cheating online. McKellar worries that entry-level students who use shortcuts to AI will lose the skills they need for advanced courses.

“I’m hearing the same thing from my colleagues in the humanities: go back to using the papers (the blue papers),” McKellar said. In addition to requiring her introductory course students to write their code by hand, paper-and-pencil tests will account for a higher percentage of grades this fall, she said.

Ronan Takizawa, a sophomore at Colorado College, had never heard of "exam books." As a computer science major, it felt like a step backwards for him, but he agreed it would force students to learn the material. "Most students are unreliable in discipline and still use ChatGPT," he said. Paper-and-pencil tests "really force you to understand and learn concepts."

Manchuan says students are sometimes confused about when AI is acceptable and when it counts as cheating. Using ChatGPT to help with certain assignments, such as summarizing reading, doesn't seem to be any different than YouTube or other sites that students have been using for years, he said.

Other students said the arrival of ChatGPT made them worry that they were being accused of cheating instead of cheating.

Nathan LeVang, a sophomore at Arizona State University, said he is now double-checking all his assignments by using artificial intelligence checkers to protect himself.

In a 2,000-word essay, the detector flagged certain passages as "written by 22% humans, mostly voiced by artificial intelligence."

"I was like, 'That's absolutely not right, because I just sat here and wrote it word for word,'" LeVine said. But he still rewrote those passages. “It’s okay if I spend an extra 10 minutes testing after I’ve written the article to make sure everything passes. It’s extra work, because I think it’s an unavoidable fact that we’re going to be tested.”

(Compilation: Chen Bingxuan)

(Editor in charge: Jiang Qiming)

(Source of the article: New Sancai first release)