PN24: Assessment: Practical Work – in the age of GenAI

We set practical work because it claims something no other form of assessment can: that this student can do it safely and competently, with their own hands. The bench itself is largely beyond the reach of an AI tool. The trouble is that we rarely mark the bench. We mark the pre-lab quiz, the report and the reflection, and an AI tool can now supply almost all of them. Nikolic et al. (2025) found the report was the most used form of laboratory assessment, and concluded that its use should be substantially reduced.

So the claim drifts quietly from “can do” to “can describe”. Watching every student every week isn’t the answer either, since unrecorded observation creates a reliability problem of its own.

This note argues for something more modest: choose the two or three moments where the doing matters most, and assess them directly, briefly and well. Mann et al. (2025) show how a short observed station can replace a written report at comparable workload, and Agustian et al. (2022) explain how we drifted away from assessing performance in the first place. A first-year chemistry vignette shows the redesign at work, burette and all.