Over the past two decades, education at all levels has increasingly incorporated collaborative learning. However, when asked about experiences with (computer-supported) collaborative learning, students typically mention (a) unequal participation—up to free-riding, and (b) dissatisfaction with the assessment process and outcome. Moreover, teachers often express they find assessment of collaborative learning challenging. In this talk I will address the position of assessment within an overall perspective on the design of collaborative learning, as well as issues regarding construct validity and the degree to which the assessment of collaborative learning is (mis)aligned with student behavior from the assumption that an assessment ought to reflect students’ individual domain-specific ability. Finally, I will address challenges and potential solutions for assessment of (computer-supported) collaborative learning. In particular I will highlight the role and added value of peer assessment and peer feedback for assessment of collaborative learning, and the potential added value of technology to aid assessment of collaborative learning. It is essential that we deepen our understanding of methods to assess collaborative learning, as well as the implications (intended or unintended) and their limitations, to enhance teacher, student and researcher awareness for as well as the need to develop ‘collaborative learning assessment literacy’.

Large-scale assessments such as PISA have shaped educational research and policy for over two decades. Yet as AI becomes embedded in classrooms, a critical question emerges: what can this evidence base tell us about the assessment of student learning, and where does it fall short?

Drawing on twenty years of research across Oslo, Oxford and Melbourne, I argue that large-scale data offers powerful insights into equity and learning patterns but must be combined with new studies co-designed with teachers and students to capture how students engage with AI in real classrooms. PISA questionnaire scales are vulnerable to validity threats, a challenge that intensifies when measuring emerging constructs such as AI literacy and self-regulated learning in digital environments. Classroom evidence offers what large-scale data cannot.

With PISA 2025 results due in 2026 and PISA 2029 introducing Media and AI Literacy as its innovative domain, the field faces a defining moment. Sustainable, equitable assessment in the age of AI requires both kinds of evidence and the will to build them together.

Assessment and feedback scholarship is replete with exciting innovations, with our SIG being a wonderful reminder of this. But what about grades? Grading systems have remained stubbornly stagnant over the history of educational institutions. Marking scales have been altered, and alternative grading methods have been proposed, yet the fundamental need to provide grades persists. This is despite scholars critiquing grades’ myriad shortcomings for more than a century. For better or worse, grades are now part of the fabric of life in education, and the life of young citizens more broadly. This presentation asks whether it might be time to move beyond grading as we know it – and then tries to clean up the mess such a question creates.