The Sydney private school that wants real data on how AI affects children

Australian schools are flying blind into the widespread adoption of generative AI, experts have warned, and are so concerned by the lack of local research they are partnering with universities to conduct their own investigations.

Researchers warn a lack of Australia-specific research leaves teachers at risk of having to improvise without the support of a solid evidence base.

When researchers from The University of Sydney and Barker College Institute examined 271 global peer-reviewed studies on generative AI’s impact on young people, they found only four were Australian, and none examined primary school students.

University of Sydney academic and author Natasha Arthars said this meant schools, researchers and policymakers “simply don’t have a clear picture” of generative AI use and the outcomes for students.

“I’m honestly still very surprised at how little research there is here in Australia, considering how much we have in the higher education space,” she said.

While research lags, adoption has exploded. The NSW Department of Education has an in-house generative AI chatbot to guide students, and it has moved to slash take-home HSC assessments due to the risk of AI cheating.

One recent report, excluded from the rapid review as it was not peer-reviewed, warned Australia’s education system faced real and urgent risks from students using AI to cheat and avoid thinking for themselves.

A spokesperson for the NSW Education and Standards Authority (NESA) said, “robust independent data in the primary school space has been limited but is beginning to emerge”, adding its current priority was to protect HSC learning and credentialing.

The rapid review found primary school children were turning to AI while foundational skills – reading, writing, mathematical reasoning, scientific thinking, metacognition and self-regulation – were developing.

Barker College Institute director and teacher Dr Matthew Hill said Australian research in primary and high schools was vital because international studies yielded contradictory results.

“When we say ‘it depends’, that sounds like it’s a cop-out, but actually that’s an answer that elevates the importance of the role of the teacher in making good decisions,” Hill said.

Arthars said: “AI is not inherently good or bad for learning.” Rather, it hinged on three factors: student characteristics, the specific tool used and the cultural context of the education system.

For instance, quiet students less likely to ask questions are turning to AI at a rapid rate.

“On one hand, that’s really encouraging to know that those students are now able to get support,” Arthars said. “But we don’t know what support they’re getting and whether it’s actually helping them.”

Students were also more likely to trust anthropomorphic tools that mimicked teachers over a simple search bar. “If they see it as a source of authority, that’s problematic,” Arthars said.

The review found the use and impact of AI varied greatly among countries: Chinese students were more likely to trust AI responses compared to Western nations.

“The school system in the USA versus the school system in Australia is going to look very different, and that is going to influence outcomes,” Arthars said.

The rapid review also found AI’s benefits to be paradoxical: while generative tools were often associated with increased interest, motivation, engagement and reduced anxiety, this increased engagement did not necessarily translate to better understanding.

Instead, tools could encourage dependence and lead to students trusting the AI without verification.

“The key implication is that decisions should focus not only on whether GenAI is used, but also on how all other conditions surrounding its use are configured and how they interact,” read the report.

A spokesperson for NSW Education Minister Prue Car said the government was actively considering measures to prevent AI from replacing students’ own thinking and authorship.

“We will continue to listen to experts as this rapidly evolving technology develops and do what it takes to protect student learning and cognitive development, regardless of age.”

Across the country, academics, experts and researchers have called for a unified approach to AI, under the banner of the Castlereagh Statement.

The statement, backed by more than 60 experts, described Australia’s overall response to generative artificial intelligence in education as “fragmented, stalled and at odds with national plans to improve our social cohesion and economic productivity”.

“Our education and training systems were designed for an era when information and cognitive labour were scarce,” the statement read.

“That era is over. Generative AI is matching or exceeding human performance on many of the cognitive tasks our systems were built to develop and assess, and it’s advancing faster than our institutions can respond.”

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.

You have reached your maximum number of saved items.

Remove items from your saved list to add more.

License this article

Emily Kowal is a Walkley award-winning education reporter for The Sydney Morning Herald.Connect via email.AdvertisementAdvertisement