USyd releases AI green paper, UQ questions ed-tech evidence base

Digital Attitudes, 12 August 2026

Wednesday, and three separate Australian voices this week are asking the same question about school and university AI, which is how anyone is going to know what actually worked.

USyd's AI green paper: this is not an AI strategy, it is an education strategy for an AI-infused world

The University of Sydney released Education in the Age of AI on 4 August, a green paper authored by Adam Bridgeman (DVC Education) and Danny Liu (Cogniti lead) that reframes the university's existing graduate qualities of judgement, ethical reasoning, creativity, agency and active citizenship as more important, not less, as AI capability rises (Teaching@Sydney summary of the green paper). The paper is explicit that it is not being pitched as an AI strategy, it is being pitched as an education strategy for an AI-infused world, and it lays out five interconnected domains for USyd to work on: student capability, learning experience, assessment practices, educator capability, and student experience. A final strategy is due in November 2026, and it builds on the Sydney Assessment Framework and the university's Cogniti platform.

Why this matters: this is the second Group of Eight signal in ten months (after the September 2025 Sydney Assessment Framework) that USyd is going to move first on what a post-AI degree looks like at scale in Australia. The framing is the interesting part. By deliberately refusing to call this an AI strategy, USyd is trying to head off the AI-strategy-as-appendix problem that has left most Australian universities with a policy document that no one uses. For every other DVC (Academic) in the country, the practical question in the next quarter is whether their institution can write a comparable document by end of Q4 2026, because the receiving employers are already starting to read it that way. For school leaders, the paper is a preview of what the marking end of Year 12 is going to look like within two years: assessment that grades the process a student can defend, not the polished output an AI can produce.

UQ's Jason Lodge: the ed-tech evidence gap in Australia is bigger than any of the policies

Professor Jason Lodge, director of the Learning, Instruction and Technology Lab at UQ, published a substantial opinion piece in The Conversation on Monday arguing that federal Education Minister Jason Clare's recent worry about too much technology in schools is missing the harder question, which is that Australia has almost no evidence infrastructure to tell whether any of it is working (The Conversation, 10 August). Lodge's numbers are pointed. Only one in five ed-tech products in US classrooms has any evidence of improving learning. The Rudd and Gillard governments spent $2.4 billion on the Digital Education Revolution laptop rollout between 2008 and 2013 without measuring outcomes. Only 50 per cent of Year 6 students and 37 per cent of Year 10 students met national digital skills standards in the 2025 NAP-ICT sample. Australian teachers already work a 46.5 hour week at elevated burnout, and no one currently knows whether adding an AI layer helps or hurts.

Why this matters: the piece is a shot across the bow of the National Cabinet AI standards work (see the 10 August brief). Lodge is naming the risk that every state buys its own purpose built AI (SA EdChat, NSW EduChat, Qld Corella, WA ClassmAIte) without a shared measurement layer, and the sector ends up with four expensive tenanted stacks and no way to compare their effect on learning. AERO's current evidence remit is not structured to answer that question at national scale, and the KPMG governance review (see 15 May brief) is going to determine whether the incoming Teaching and Learning Commission inherits the mandate. For school and university leaders, the read is that the burden of proving what your AI stack is actually doing is going to shift from the vendor onto you, in writing, inside twenty four months.

AI in Education Conference lands at Nyaal Banyul, Geelong on Thursday

The Informa AI in Education Conference for Victoria runs at the Nyaal Banyul Geelong Convention and Event Centre on Thursday 13 August, with the program built around AI in classroom practice, personalised learning, and assessment reform (Informa event page). Speakers include Nick Gissing (Head of AI and Innovation, Wodonga Senior Secondary College), Rachelle Crameri (Director of AI and Digital Innovation, Ivanhoe Grammar School), Jack Leask (Monash Tech School), Leah Li (Senior Lecturer, RMIT), and Trudy Purcell (Head of House and Commerce teacher, Geelong Grammar). Purcell previewed her session in EducationHQ on Monday, arguing that assessment now has to assume AI has been used and grade the process the student can defend rather than the polished output the model can produce (EducationHQ preview).

Why this matters: the Geelong room is where the Victorian government sector, the independent sector, and the vendor floor line up on the assessment question the same week NSW's Prue Car has NESA looking at a moratorium on unsupervised HSC take home tasks (see yesterday's brief). If the Victorian DET wants to keep secondary school assessment reform in front of the NSW process, the practical work is in what gets said on stage tomorrow. Note the sponsor list (Toddle as Gold, Thinking Mode as Supporting) because that is where the next twelve months of Vic classroom AI procurement gets negotiated. For principals not attending, the published speaker list is the read of who the sector currently thinks is worth listening to on this question.

Three voices, one question, and no one on the receiving end of a graduating student is going to accept "we ran some tools and it felt good" as a full answer by the end of next year. More on the operational landscape at digitalattitudes.com.au.