When done well, providing feedback is one of the most powerful strategies teachers can use to support student learning. As artificial intelligence tools and technologies evolve rapidly, schools and education systems are grappling with the challenges and opportunities they present.
In student assessment, AI offers new possibilities for feedback while also raising important questions about quality, accuracy, teacher oversight and, crucially, whether the feedback improves the learner, not just the piece of work they’ve been tasked with completing.
Building an evidence base of effective practice amid fast-evolving technology requires researchers working in close collaboration with teachers and students – that was one of the key messages from a recent webinar exploring the implications of AI for assessment feedback in schools. Hosted by the Australian Council for Educational Research (ACER), the session featured expert insights from Dr Fabienne van der Kleij and Professor Therese Hopfenbeck.
Debunking common assumptions about feedback
Dr van der Kleij, Senior Research Fellow in Assessment Transformation at ACER, said the conversation about integrating AI tools needs to start with our understanding of feedback itself.
‘We know from research that feedback can improve learning if it triggers thinking, for example, through changes in knowledge, skills, or strategies, feelings and actions that then lead to changes in what students do next,’ she told the audience of predominantly K-12 teachers and leaders. ‘Feedback is one of the most widely researched topics in education, and even though we know it can be effective, in practice it is often not.’
Dr van der Kleij debunked 3 common assumptions about feedback.
- Features of the feedback message determine its effectiveness. ‘Research shows that the same feedback can have very different effects on different learners.’
- More feedback leads to better learning outcomes. ‘Sometimes less feedback, or less specific feedback, is more powerful when activating learners. It can also create dependency – if students become increasingly reliant on feedback from others to evaluate and improve their work, we might be undermining the very self-regulation that feedback is meant to develop.’
- Students pay attention to feedback and know how to use it. ‘Students might not pay attention to feedback, they might not understand it, or they might be willing to use it but don’t know how or make superficial changes. And we see that a lot in our research in the context of writing, for example.’
Webinar attendees were urged to first think about the fundamental purpose of feedback – improving the learning, not what students produce. ‘If feedback focuses on fixing students’ work for them, then we’re missing the point. … if our underlying model of feedback is flawed, AI simply helps us make problematic practices more widespread.’
How AI might shape assessment and feedback
Moving onto how AI tools and technologies have the potential to shape assessment and feedback in education, Dr van der Kleij shared 3 exciting opportunities: the capacity to provide instant and on-demand personalised feedback to learners; tools offering insights for teachers into patterns in student behaviour and learning, which could then inform next teaching and learning steps and targeted support; and new ways of learning with AI – an important area if teachers are to prepare students for a future where these technologies will be commonplace.
As with the adoption of any new technologies in education, realising the benefits requires careful attention to its challenges and risks.
The first consideration for teachers and leaders, Dr van der Kleij shared, is ensuring the use of AI tools is purposeful. Again, as with any technology use, look past the hype and think about why you want to use it in the classroom. ‘Don’t just deploy AI because we’re afraid to get left behind.’
Be mindful of the limitations of tools in terms of feedback quality and accuracy. For example, is it providing feedback on the focus area that we want it to? There are also ethical considerations around student data and privacy.
Dr van der Kleij reiterated that more feedback isn’t always more effective. ‘Some AI tools provide instant corrections, they might rewrite students’ work, or they might optimise the product, but how much does a student actually learn from this? For example, in the context of writing, research shows that students tend to make mostly superficial changes in response to feedback. Generative AI doing the writing for them poses a huge risk of exacerbating this problem.’
Teachers also need to have oversight, and there’s limited research in relation to this and meaningful integration of AI tools into assessment and feedback in the classroom.
Critical student capabilities
Webinar attendees also heard about recent research exploring assessment feedback in education. Dr van der Kleij shared details of a project in partnership with the International Baccalaureate that reviewed AI feedback in literacy and found 3 critical capabilities.
- Evaluative judgement. ‘Students need to critically assess the quality, accuracy and relevance of AI-generated feedback.’
- Feedback seeking. ‘Students need to actively elicit meaningful feedback. For example, through effective prompting and questioning.’
- Agency and ownership. ‘They need to be able to decide how to use feedback over time, maintaining ownership of their work and learning.’
A key takeaway is that these capabilities must be deliberately developed, especially in schools.
Sharing examples from the classroom
Professor Therese Hopfenbeck, Director of the University of Melbourne’s Assessment and Evaluation Research Centre (AERC), also spoke about the importance of teacher guidance and modelling best practices to students.
She is currently working on a 5-year TRAINE (Tracking AI in Education) study with Principal Investigator Dr Samantha-Kaye Johnston, collecting data on how Generative AI tools are influencing teaching, learning and assessment in Australian secondary school classrooms. The first results are due to be published in June.
Both experts highlighted the importance of sharing and learning from effective classroom practice, and researchers working with teachers and students to build the evidence base.
News and insights on AI in education research and practice
Over the coming weeks and months, Teacher will be sharing examples of how teachers and education systems are trialling and using AI tools in their contexts.
We’ll also be speaking to Professor Therese Hopfenbeck about the challenges of using AI in assessment feedback and the solutions it offers, the initial results of the TRAINE study, and examples of effective practice emerging in Australian schools.
Sign up to the Teacher bulletin to stay up to date with the latest. If you’d like to share what’s happening in your school, email teachereditorial@acer.org and we’ll be in touch.
Further reading
Brooks, C., Burton, R., Van Der Kleij, F., Carroll, A., Olave, K., & Hattie, J. (2021). From fixing the work to improving the learner: An initial evaluation of a professional learning intervention using a new student-centred feedback model. Studies in Educational Evaluation, 68, 100943. https://doi.org/10.1016/j.stueduc.2020.100943
Chen, Z., Zhu, J., McGrane, J., Hopfenbeck, T. N., Wang, Y., & Ji, Y. (2026). More inspiration and attention: How Generative AI tools impact graduate students’ affective engagement in L2 source-based academic writing. Computers & Education, 242, 105495. https://doi.org/10.1016/j.compedu.2025.105495
Crompton, H., Edmett, A., Ichaporia, N., & Burke, D. (2024). AI and English language teaching: Affordances and challenges. British Journal of Educational Technology, 55(6), 2503–2529. https://doi.org/10.1111/bjet.13460
Gamlem, S. M., Johnston, S.-K., Moltudal, S., McGrane, J., & Hopfenbeck, T. N. (2026). Generative AI in initial teacher education: Exploring the alignment of perceptions and experiences of pre-service teachers and their teacher educators. Teaching and Teacher Education, 172, 105382. https://doi.org/10.1016/j.tate.2026.105382
Gamlem, S. M., McGrane, J., Brandmo, C., Moltudal, S., Sun, S. Z., & Hopfenbeck, T. N. (2026). Exploring pre-service teachers’ attitudes and experiences with generative AI: A mixed methods study in Norwegian teacher education. Educational Psychology, 46(1), 27–51. https://doi.org/10.1080/01443410.2025.2528663
Hopfenbeck, T. N. (2023). The future of educational assessment: Self-assessment, grit and ChatGPT? Assessment in Education: Principles, Policy & Practice, 30(2), 99–103. https://doi.org/10.1080/0969594X.2023.2212192
Hopfenbeck, T. N., Zhang, Z., Sun, S. Z., Robertson, P., & McGrane, J. A. (2023). Challenges and opportunities for classroom-based formative assessment and AI: A perspective article. Frontiers in Education, 8, 1270700. https://doi.org/10.3389/feduc.2023.1270700
Lipnevich, A. A., & Smith, J. K. (2022). Student – Feedback Interaction Model: Revised. Studies in Educational Evaluation, 75, 101208. https://doi.org/10.1016/j.stueduc.2022.101208
Milano, S., McGrane, J. A., & Leonelli, S. (2023). Large language models challenge the future of higher education. Nature Machine Intelligence, 5(4), 333–334. https://doi.org/10.1038/s42256-023-00644-2
Molenaar, I. (2022). Towards hybrid-human‐AI learning technologies. European Journal of Education, 57(4), 632–645. https://doi.org/10.1111/ejed.12527
Saubern, R., Taylor-Guy, P., & Van Der Keij, F. (2022). Introducing the Education Technology Value Evaluation Tool for Schools. 2022 International Conference on Assessment and Learning (ICAL), 1–6. https://research.acer.edu.au/ical/16
Yang, S., Taylor-Griffiths, D., van der Kleij, F., Taylor-Guy, P., & Saubern, R. (2025). Empowering teaching and learning with educational technology: Literature review. Australian Council for Educational Research. https://doi.org/10.37517/978-1-74286-784-7
When you give feedback, what is your primary goal? How do your students typically respond to feedback? What evidence do you have that they understand and use it?
This article discusses 3 critical student capabilities in relation to AI feedback. On your own, or with colleagues, choose one and unpack what it might look like in practice to intentionally teach and deliberately develop this capability in your context.
