One of the learners I remember best earned credit for a life spent on the land — trapping, harvesting, the knowledge a written test would never see. Another was a parent whose years of raising children were, at last, read as learning rather than as time away from it. In prior-learning assessment you do not ask people to sit a test that pretends they know nothing. You ask them to document what they can already do, and you assess the evidence and the judgment behind it.
That is the whole model, and it was AI-proof before AI existed. Recognition of prior learning never assessed a product a machine could fake. It assessed documented process, real artifacts, and the reasoning that ties them to a standard.
Evidence, not gaps
The move that makes it work is a stance: read a life as evidence, not as a deficit. A portfolio is a designed argument — here is what I did, here is what it demonstrates, here is how it maps to the outcome. Experiential learning has to be turned into a claim and defended, which is precisely the cognitive work AI cannot do on the learner’s behalf (Kolb, 1984).
Prior-learning assessment survived AI before anyone needed the phrase. It assesses documented process and judgment, not a product.
What every assessor can borrow
You do not need a formal recognition mandate to use its logic. Ask for the portfolio, not just the product. Ask the learner to show the decisions, name the standard their evidence meets, and defend the fit. I have watched the moment a learner realises their own experience counts — it changes how they show up for everything after. That design is honest, humane, and quietly robust against shortcuts.
Why AI raises the stakes
As product-based assessment collapses under AI, the practices that already assess process become the model rather than the exception. The returning adult with a portfolio is not an edge case to accommodate. They are the template for assessment that still means something.
The move
Add one portfolio element to a course that has none: a short, evidenced claim the learner must defend. It is the smallest step toward assessment that AI cannot hollow out.