AI and the learning alignment problem

2 min read

Presentation

GRNEN AI Webinar Series · Global Research Nursing Education Network · 11 June 2026

Invited keynote webinar

Overview

An invited keynote for the Global Research Nursing Education Network’s AI webinar series, delivered to nursing educators working across the US, Canada, and Africa.

The talk borrows a term from AI safety. The value alignment problem is the difficulty of specifying what you actually want a system to optimise, rather than a measurable stand-in that comes apart from it under pressure. Education has the same problem and has had it for far longer: we cannot observe learning directly, so we assess proxies — artefacts, grades, submissions — and then manage the proxy as though it were the thing. Generative AI didn’t create that gap. It made it impossible to keep ignoring, by severing the link between the work of becoming a nurse and the evidence of that becoming.

The session covers:

  • Why learning is unobservable, and what follows from assessing proxies for it — including Goodhart’s problem, where a measure that becomes a target stops being a good measure
  • The four premises of nursing formation, and what each one requires that an artefact cannot evidence
  • Why the two responses so far — better detection, and discursive responses like policies and AI declarations — cannot address a structural problem
  • What educators contribute once AI agents raise the ceiling on what students can produce, and how to design for the conditions that support cognitive struggle rather than for the artefact that used to stand in for it
  • The equity dimension, taken seriously for an audience spanning very different contexts: a design that assumes every student arrives with capable agents, connectivity, and time builds existing advantage into its foundations

It shares its underlying argument with What is the work?, delivered to the RCN Education conference in April, and reframes it through the alignment parallel for a research-nursing audience.

Slides

Use arrow keys or click to navigate slides. Open slides in full screen

Continue reading

  • What is the work?

    A keynote for the Royal College of Nursing's Education conference. Examines how generative AI has severed the inferential chain between assessment artifacts and the learning they were meant to evidence, and what nursing education needs to do structurally in response. The argument moves from the current AI landscape, through the premises of nursing formation, to why discursive responses (policies, declarations) cannot address a structural problem.

  • The verification trap

    Institutions everywhere now insist on the same rule — use AI, but always verify its output. Yet verifying AI output is hard, slow, and often inconclusive, even in domains where you're competent to judge it. This post pushes that advice to its limit to see what breaks, and finds that as models improve, the verification mandate quietly trains the opposite of the scepticism it intends.

  • What software engineers can teach us about AI in doctoral research

    PhD students are using AI across their doctoral work, but current policies focus on permission rather than specification. Drawing on how software engineers solved a structurally identical problem, this post introduces the research harness: a seven-part operating context that makes AI contributions visible, traceable, and supervisable. The harness addresses drift, offloaded thinking, and attribution failures — not by restricting AI use, but by specifying what the agent can and cannot do within the work.