AI – keeping it helpful (and harmless) in education

'We're in the beginning of a new era; AI is coming in education, so let's do it in a way where all children get access to knowledge on how to use it for learning.’   

Professor Therese N. Hopfenbeck delivered this call to action at Research Conference 2025. Her presentation to educators, researchers, policymakers, school and system leaders had included several audience discussions and the enthusiasm had been hard to quell each time. 

The Director of Melbourne University’s Assessment & Evaluation Research Centre was speaking on ‘AI-enhanced assessment: From theory to evidence-based practice’.  

Assessment is a subject she has considerable expertise in; she has been an advisor on the implementation of formative assessment programs in India, South Africa, Tanzania, Norway and the Emirates. She has also been involved in policy work for UNESCO, the OECD and the Norwegian Ministry of Education, and led the Oxford University Centre for Educational Assessment for 7 years. 

In every assessment, the focus should remain on knowing how students learn, keeping them at the centre of every approach, and creating rich tasks for deep learning, Professor Hopfenbeck told her audience. 

AI-enhanced assessment could bring benefits for personalised learning, including through adaptive testing that adjusts difficulty based on student responses and the ability to track individual student progress over time.  

The skills teachers need now, Professor Hopfenbeck said, include AI literacy, critical thinking and the ability to navigate uncertainty.

For students, it was important they learn to:  

• be honest about which models they’ve used,  
• explain what it was used for, 
• explain how the information has been verified and 
• highlight where in their assignments they have used AI 

Professor Hopfenbeck referred to work that she and her team had done on AI and the challenges and opportunities for classroom-based formative assessment.  

Feedback, which was key to successful assessment, could be enhanced by AI to help student literacy, Professor Hopfenbeck said.  

Speaking during a break at Research Conference 2025, she shared an example based on findings from Norway. 

‘The Norwegian Research Council has funded some really interesting studies looking into how teachers have used AI tutors providing feedback to students when they’re trying to improve their texts,’ she said.  

‘Having studied formative assessment for so many years, I know that teachers generally say that’s really hard and it’s getting harder the more students you have in a classroom. 

‘It was very encouraging to see that this project, conducted over many years, turned out to be a project where teachers said “This is great; I would never have been able to provide that much rich feedback to the texts” and the students were also very impressed with how they improved the text.’ 

A panel discussion at the conference attracted similarly high levels of interest from participants in introducing AI effectively.  

Facilitated by Dr Fabienne van der Kleij, the principal researcher responsible for ACER’s work in school and system improvement, the panel brought together diverse perspectives.  

Experts included ACER’s Dr Tim Friedman, who has worked internationally on projects like preparing for life in a digital world and Dr Suijing Yang, whose work on AI includes ensuring teacher agency in a technology-empowered world.  

Also joining Professor Hopfenbeck on the panel was Dr Kelly Stephens from the University of Technology Sydney, who has written on securing digital equity in Australian education and helped shape how we think of AI. 

Considerations in using AI

Panellists raised the following issues for consideration:  

• Access and equity – Does every school have access to safe programs? 
• Data - Is the input representative of student diversity and does the outcome avoid bias? 
• Design - Has the development of a program or app been market-driven by tech companies interested in quick returns? 
• Knowledge – Do teachers have the tools to know when and how best AI can be used? 
• Awareness – Are students aware of the danger of providing personal details and data when information is requested from them to subscribe or use? 

New resources  

Dr van der Kleij and Dr Yang, together with Jonathan Brazil, a PhD student at the University of Queensland who undertook an internship at ACER, have recently published guiding principles and practical examples for using AI in teaching and learning. 

Dr van der Kleij believes the work is both timely and useful for those looking for a substantial response to the ‘hot topic’ of including AI in teaching and learning.  

The work identifies a ‘PATH’ forward, through the following actions: 

• Promoting teaching and learning
• Advancing wellbeing and social interaction
• Training for AI literacy
• Harnessing AI ethically 

ACER also offers professional learning for teachers and school leaders, with modules including fundamental concepts and applications of AI in learning and teaching, ethical considerations and limitations, human-centred approaches to AI integration and practical skills for planning, monitoring, and evaluating AI use in the classroom.. 

It aims to build teacher confidence in their investigation and use of AI as well as the ability to find answers to questions about AI as they arise. 

Working together for the best outcomes 

During her presentation, Dr Hopfenbeck shared a conversation with an AI assistant as an example of what teachers can do with their students to focus on the limitations and challenges of using AI. 

She had asked Claude ‘How can I know if a text has been written by Claude?’ 

The AI had apologised, noting 'I’m Claude, an AI assistant created … to be helpful, harmless and honest; I don’t have information about my own capabilities compared to other AIs or humans.'  

The eagerness to uncover the benefits of AI in education shown at Research Conference 2025 suggests that the human task of gathering that information is well in hand.