Subscribe
SUs
SUs

Wonkhe SUs · The 2026 webinar box set

The robots are here

What AI really means for students, assessment and learning – and why the answer is about assessment design, not banning the bots.

Checking your subscription…

One moment while we confirm your access to this recording.

Mack Marshall runs through everything SUs need to know about AI in higher education: the scramble from "ban it" to today's principles-based policies, why framing AI policy around academic misconduct misfires, and what the OIA and OfS have said. The bulk draws on Wonkhe's "Trained to Stop Learning" research into AI, students, assessment and learning – why students are increasingly submitting work they can't fully explain, and why the fix lies in assessment design rather than detection.

What we cover

  • The policy scramble – how universities lurched from "ban it" in 2022/23 to broad, principles-based AI policies, nearly all framed around assessment and academic misconduct.
  • Why misconduct-framed policies misfire – students use AI far beyond writing essays, so a policy built around assessment leaves everyone unsure what actually counts.
  • Academic misconduct – the OIA's case note favouring educational over punitive responses, and why the burden of proof sits with the provider (and complaints succeed when it isn't met).
  • The external picture – the post-16 white paper directing OfS to examine AI, and employers wanting AI skills alongside communication and interpersonal strengths.
  • Three groups of students – those too scared to touch it (and falling behind), those who use it badly at 3am and get caught, and those who use premium tools well and don't – and the access and participation gap that opens up.
  • The institutional picture – AI literacy, assessment redesign and tool deployment as top priorities, but policy too often mistaken for practice.
  • Trained to Stop Learning – Wonkhe's research (1,000+ students across 52 providers, plus focus groups) on whether students actually learn when they produce work.
  • The understanding gap – 47% worry grades don't reflect what they know, 38% submit work they couldn't explain without their sources, and courses that reward outputs see AI used at double the rate.
  • AI isn't one thing – six modes from search-engine replacement and scaffolding to debugging partner, always-on tutor and production accelerator, only the last of which is really an integrity concern.
  • It's assessment design, not AI – accountability moments, the equity and disability dimensions, peer learning, and why the sector should start with what assessment is for, not what to do about AI.

Things to read

← The 2026 webinar box set