Uni student Daniel hesitates when asked if he has used ChatGPT to cheat on assignments before.

His answer is “no”, but the 22-year-old feels the need to explain it further.

“I don’t think it’s cheating,” he said.

“As long as you accredit it and use it for like a foundation for your assignment I think it’s fine.”

Daniel says he has used ChatGPT in his studies.(BTN High: Cale Matthews)

Schools and universities have been scrambling to keep up since ChatGPT and other generative AI language programs were released in late 2022.

University student Lan Lang, 18, said quite a few people used generative AI for assessments such as English assignments.

“I do get Chat to like explain stuff to me if teachers don’t really explain it that well,” Lan Lang said.

Sam and Lan Lang say generative Al can be a helpful tool for students.(BTN High: Cale Matthews)

She said she used AI detection software on her work.

“We put it through Turnitin, which just basically detects if you’ve used AI, or if you’ve copied off anyone else’s work,” she said.

Caught out in schools

High school teacher Ryan Miller said he wasn’t seeing a lot of generative AI used in the Year 12 and Year 8 classes he taught but understood from colleagues other age groups were using it.

Ryan Miller says students get a warning when caught using AI for assignments.(Supplied: Ryan Miller)

“What I hear, when I’m in the staff room, is that a lot of Year 9s, 10s, [and] 11s are pushing the boundaries,” Mr Miller said.

He said Year 12 students tended to be more careful after being warned at the start of the year and constantly reminded of consequences.

“Basically, they’re told if their work is seen to be made … predominantly with AI, that it won’t be assessed,” he said.

Mr Miller said Year 8s, being a little newer to the school, hadn’t used it as much.

He said teachers tended to give students a warning if they were detected using generative AI.

“And nine times out of 10 they’ll probably own up to it and say, ‘Yeah, look, it wasn’t … 100 per cent my own work’,” he said. 

He said students would rewrite the work so it could be assessed again.

“But it’s sort of a one warning per kid, per year for most teachers, I think,” he said.

Fellow teacher Hugh Kinnane said generative AI was probably “pretty rife” in assignment work.

He said he most regularly saw it cropping up with students who were trying to avoid doing any work.

“And then it’s a last-minute job,” he said.

Jennie Shaw says her university has embraced the use of artificial intelligence.(BTN High: Cale Matthews)

Drawing the line

University of Adelaide Deputy Vice-Chancellor Academic Jennie Shaw said while her university embraced the use of AI, it could still be used to cheat.

“So we’re saying, of course, that is not allowed,” Dr Shaw said.

She said generative AI was included in academic integrity modules for first-year students.

“We make it really clear to students what is OK and what is not OK,” she said.

Dr Shaw said there were instances when students were encouraged to use generative AI and then critique the quality of its answer.

“What we are asking our students and our staff to do is to reference when they do use it,” she said.

She said it was a requirement that as much content as possible was checked by similarity detection software.

According to Turnitin’s website — which is used by the University of Adelaide as well as many other universities across Australia to detect AI-generated content— the company is committed to a false positive rate of less than 1 per cent to ensure not students are falsely accused of misconduct.

AI arms race

The software has put students at the centre of a battle for superiority between programs generating answers for their assignments and those designed to catch them out.

And according to Australian Institute for Machine Learning senior lecturer Feras Dayoub, some are getting caught in the crossfire.

Feras Dayoub says AI detection tools are not 100 per cent accurate.(BTN High: Cale Matthews)

He said companies that created AI chatbots were trying to be undetectable while companies that created AI detection software wanted to detect everything.

“There will be a lot of false positives,” Dr Dayoub said.

He said it could be an unpleasant experience for the student if the detector was wrong.

Students Ethan (left) and Nicholas say they have noticed issues with AI-detecting programs.(BTN High: Cale Matthews)

University student Ethan, 19, said single words were sometimes highlighted in his Turnitin submissions.

“It can be a bit inaccurate,” Ethan said.

Dr Shaw said she understood the detection software had its faults.

“We would find probably two thirds of anything they pick up saying there’s some unacceptably high levels of similarity here is often just picking up patterns in language,” she said.

“I know some universities have chosen to turn it off because it does turn up lots of false positives.

“We’re choosing to use it at this point.”

Changing education

The Department of Education released a nationwide framework in December last year for the use of generative AI in schools.

Dr Shaw said the technology was changing the way teachers taught and students learned.

“But we still need students to have deep knowledge,” she said. 

“We need them to know how to use the tools in their profession. 

“And again, one of those in many professions will now be generative AI, and we need them to be able to call out when it’s wrong.”

Dr Dayoub said he would prefer a future in which there was no need for detectors because people had changed the way they taught and assessed.

He said another option would be to take a stricter approach, where students did the work themselves and there would be no help.

“In that case you need the detectors so there will be a huge market for these detectors and it will become a race,” he said.

“I don’t like that future.”