Validity and authenticity: How to treat cognition and AI
26 May 2026
In this post, CRADLE PhD student Sekartiyasa (Sky) Kusumastuti reflects on our fourth seminar for 2026, ‘Entangled intelligence? Distributed cognition, AI agents, and assessment validity’. In this seminar Professor Jason Lodge acknowledged that we don’t yet have all the answers, and challenged the audience to consider some difficult questions around assessment and generative AI.
Entangled intelligence? Distributed cognition, AI agents, and assessment validity
In his CRADLE seminar, Professor Jason Lodge from the University of Queensland invited us to look beyond the familiar crisis language surrounding generative AI and assessment. Rather than treating AI as a problem that can be solved through tighter rules, detectors, or simple traffic-light frameworks, he argued that we need a deeper understanding of the mechanisms shaping learning, cognition, and assessment practice in AI-integrated environments.
Dr Thomas Corbin introduces Professor Jason Lodge at the seminar (Siham AbuKhalaf)
A useful distinction in the seminar was between the acute and chronic problems of AI and assessment. The acute problem is the immediate institutional pressure to respond to cheating concerns, media narratives, and demands for secure assessment. Ring-fenced assessment, such as supervised exams and interactive orals, may offer a temporary response to this pressure. Yet the chronic problem is more fundamental:
What should universities teach and assess when students, teachers, and professionals increasingly work with AI as part of their cognitive environment?
The seminar also raised important questions about offloading. Some forms of offloading are reasonable and even desirable, such as using tools to reformat references. Others may be educationally risky when students outsource the core thinking that assessment is meant to develop. This makes the issue less about whether AI is permitted and more about what kinds of human judgement, disciplinary knowledge, and epistemic capability remain visible in students’ work.
For me, the key takeaway was that assessment cannot be redesigned through slogans or silver-bullet solutions. If cognition is becoming increasingly entangled with AI tools, assessment design must become more sensitive to those entanglements. The challenge ahead is not only to secure assessment, but to design valid and authentic ways of evidencing learning in a world where human and machine activity are increasingly difficult to separate.
Missed the seminar? Catch up on our YouTube channel!
About Sky
Sky Kusumastuti is a Research PhD student with CRADLE. Her research interests include GenAI and assessment in higher education. Previously Sky has worked as a language educator and curriculum designer, in particular as a TESOL educator.
Discover more from CRADLE Blog
Subscribe to get the latest posts sent to your email.
