fountain pen on black lined paper

When the Essay Writes Itself, What Is the Grade Measuring?

By Oli Huggins

Every teacher and every parent has now had the same quiet realisation. The essay, the problem set, the coding exercise, the kind of work we have used for a century to find out whether a young person has learned something, can be produced in seconds by a machine that has learned nothing at all. This is usually discussed as a cheating problem, which is the wrong frame and a distraction. The deeper issue is that the grade was always a proxy. We never actually wanted the essay, we wanted the thinking the essay was evidence of, and AI has quietly severed the two.

For most of education’s history the proxy held, because producing the work was hard enough that doing it well meant you had probably understood the material. That assumption no longer survives contact with a free chatbot. A student can submit excellent work and have learned nothing, and a marking scheme built to measure the work cannot tell the difference. The completion certificate, the grade, the finished assignment, all of it now measures access to a tool rather than capability in a person.

What remains worth measuring is the part AI cannot fake on a student’s behalf, which is performance under conditions the student cannot outsource. Can they explain the reasoning aloud, can they apply it to a problem they have not seen, can they do the thing when the tool is taken away. This is harder to assess than collecting an essay, which is precisely why it is now worth doing. The schools and training providers that move first to assess capability rather than collect output will be the ones whose certificates still mean something in 5 years.

This is the principle ExpertEdge is built on. We take books from expert publishers (MIT Press, Wiley, Sage, Packt) and turn them into courses designed to certify what a learner can actually do, not just confirm they reached the final page. Content is now close to free, so the scarce and durable thing is verified capability, and that is the asset worth building a course around.

If you set assessments for a living, the practical move for this year is small. Pick one thing you currently grade by output, and redesign it to test what the learner can do without the tool. You will learn more from that one change than from any policy on AI use.

Oli Huggins is CEO of ExpertEdge

Previous post Glion strengthens the bridge between academia and industry as hospitality jobs evolve faster than ever
Next post 91% of students use AI to study, survey finds