UCC staff told it would be almost impossible to detect students cheating with ChatGPT

The emergence of ChatGPT has prompted concern within education about its potential use as a tool for cheating, with universities in many countries encouraging staff to re-think how they structure exams to ensure students cannot rely on such technologies to answer the questions for them. File photo: AP/Michael Dwyer

Staff at University College Cork (UCC) have been told it would be almost impossible to detect if students were using the controversial artificial intelligence (AI) tool ChatGPT to write assignments.

Lecturers were told that if the AI created a draft that students then edited, it was "likely to be very difficult to detect". They were also told that relying on existing tools for detecting plagiarism would not work as such tools are not “keeping apace with ChatGPT’s ability to evade detection”.

The emergence of ChatGPT has prompted concern within education about its potential use as a tool for cheating, with universities in many countries encouraging staff to re-think how they structure exams to ensure students cannot rely on such technologies to answer the questions for them.

UCC staff were told it was not practical to try and block student access to such tools but advised to link assessment to in-class discussion, as well as video, audio, or visual presentations that AI tools couldn’t access. 

Another idea was to ask students to evaluate specific research papers that were behind paywalls and therefore couldn’t be accessed by ChatGPT. The presentation said no cases of students cheating with AI had yet been detected but that it would be “naïve” to think some hadn’t experimented with the technology.

Teaching staff were also told there were limitations with ChatGPT which could generate fictitious references, give inaccurate or incomplete answers, and only had access to what was freely available on the internet. “This is, however, a fast-moving field so we expect this type of tool to become increasingly convincing,” it said.

Lecturers were also told there was great potential in AI to help them with their work in developing draft lesson plans, writing code, or creating drafts for editing.

Internal emails also detail growing concern among staff about artificial intelligence and how it could compromise assessment of students. One email said: “There is increasing concern about the emergence of AI text generators, such as ChatGPT, which can generate a well-written essay response in seconds, often for free.” 

Another said: 

Reminds me of King Canute talking about fighting the tide.

In discussions internally, UCC staff said it was important for lecturers “not to panic”. The email said: “Stress there is (currently) no technological solutions to detect essays or code done using this system. But not to panic.” 

In another message, one professor detailed how he had asked ChatGPT to write a poem in the style of Wordsworth on the subject of motorcycles. “Five seconds later I got the poem copied below,” he said, “you have to concede it’s rather good. We really have a problem here.” 

In another message, a senior member of faculty said the only silver lining from AI was that they were “threatening the business model of essay mills”. “Sadly, we can’t uninvent these things and so we will just have to adapt to live with them,” said the email.

A staff member also said UCC would likely need to update its official risk register to include AI as “one of those existential threat moments”.

In another email, a professor warned of a technological “arms race” between AI text generators and tools intended to detect it. “[It] will make it very hard for us to accuse a student of misconduct,” said the message.

A colleague responded saying: “One almost fool-proof strategy a student can use (so student culture will quickly cotton to, or likely already has) is to submit a couple of their own compositions (which could be on any subject) to ChatGPT and then ask it to ‘write an essay [for a module] in the style of me’.

“This will produce clunky papers with idiosyncratic errors that will be effectively impossible to screen.” 

Asked about the records, a UCC spokesman said: “As with all emerging technologies, UCC is keeping abreast of the development of ChatGPT and is engaging with staff accordingly. UCC continues to actively provide advice to students on all aspects of academic integrity, including the use of AI systems.”