Back to basics: Teachers adopt old-school methods to overcome AI use

The use of artificial intelligence by high school students is now so widespread that many pupils can’t imagine learning without it, new research has found.

Professor of Education Miriam Tanti leads a team at La Trobe University that has undertaken several studies this year surveying secondary school students across two states about their use of AI in their studies.

The research found that many students report they “really struggle to imagine learning without AI”, while others admitted they now never study for exams without it.

“The risk isn’t necessarily plagiarism or cheating, the risk is that illusion of understanding,” Tanti told this masthead. “They think they know a topic because they put it in a chatbot, but they’ve bypassed the effortful process of thinking that makes learning stick. That’s the real challenge ahead of us.”

Tanti and her colleagues will release a teaching guide next month that emphasises students’ need for explicit instruction – teachers modelling tasks step by step – to show students how they should be engaging with AI for learning.

“What we’re saying is that learning happens through specific cognitive processes, by managing attention, retrieving knowledge. So gen AI is useful when it protects those cognitive processes and protects that effort,” Tanti said.

St Leonard’s College Brighton is one of the many schools experimenting heavily with several teachers at the bayside school having trained AI chatbots to provide feedback to students outside class time. Teachers then closely monitor the feedback.

“It’s like a mini teacher,” year 10 student Isabelle Skewes said of a bot developed for her economics class. “[My teacher] programmed it to sometimes even mark work with assistance, and give feedback she would give. It’s basically generated to talk like her and speak like her. I ask it for a lot of feedback.”

There is plenty of discussion about responsible use of technology at the school, too.

From year 8, students learn about the biases, hallucinations and overconfidence inherent to many chatbots to ensure they’re critical of AI output.

Isabelle, for example, wrote a history essay on the role of the sea in Viking and Polynesian civilisations, with the starting point of asking AI for help. The AI response was useful, she said, but the answer was too general and included many stereotypes. She then built on the response herself and was marked on her own additions.

“The teacher turns it back on them and says, ‘Right, OK, it’s given you this answer, so what else can you do to try and verify that answer? What else can you do to criticise this answer?’” St Leonard’s director of digital innovation, Charles Neave, said.

How students write an AI prompt is another area that teachers are testing students on. “We’re showing them that they should actually get extensive answers that look at a lot of different aspects of the question,” Neave said. Some educators are thinking about tasks that grade students on the pathways they sent chatbots down and the counter-prompts they corrected course with.

However, uptake of the rapidly developing technology in schools has experts worried that students relying on AI are bypassing the cognitive benefits of traditional learning, as the fightback against its dominance gets under way.

The Victorian Education Department is taking a cautious approach. It doesn’t prescribe a single platform for schools, but is monitoring platforms as it prepares to publish guidance later this year that will address emerging risks such as over-reliance in student learning.

John Paul College in Frankston was among the early adopters of using chatbots in class in 2023; the technology has now changed the expectations of academic integrity for assignments done without supervision.

“We’ve said to our teachers, the minute you allow a student to do any work, either unsupervised in your room or you’ve asked them to do the work at home, you have to assume AI is being used,” said deputy principal Michael O’Keeffe.

Generative AI – which students may use to create “original” content such as essays, images or reports – is also forcing educators to think carefully about protecting the process of learning, with some schools doubling down on the use of pen and paper.

Haileybury uses a traffic light system for the tasks that students use AI for – a green light for an artwork might mean it can be entirely designed using AI, for example, or an essay marked yellow limits it only for brainstorming. This year, students returned to using pen and paper for most middle school exams, partly due to the risk of them using AI to generate answers during an assessment.

“You’re a bit more bulletproof when you’ve just got pen and paper,” said the school’s deputy principal of teaching and learning, Anna Sever.

“That’s why we are so focused on foundational skills. We want our students to be creative, we don’t want them to offload at those critical junctures with their learning, and we don’t want them to underestimate the value of hard work.”

Oliver Schreurs, a year 11 student at Haileybury, said he used AI by giving it the marking rubric for assessments and asking it for feedback on his work.

“AI can give us feedback on that and tell us how to restructure things or improve our work, and I think using it like that as a tool is good, while still keeping your work as your own.”

Still, even his computer science classes now have more supervised handwritten assessments.

“I personally still believe in pen and paper,” Schreurs said. “Not just to ensure the work is original, but I also think ideas can be expressed in the most original and creative way on paper, and it helps drill ideas and theories into minds instead of just typing it.”

Start the day with a summary of the day’s most important and interesting stories, analysis and insights. Sign up for our Morning Edition newsletter.