Margaret-Anne Storey recently flagged measurement of cognitive debt as an open research question. We have plenty of metrics for delivery speed — DORA, cycle time, merge frequency — but nothing that tells us whether the team that shipped a feature actually understands it.
This article proposes a concrete starting point: the Feature Comprehension Score. At the close of each feature, an AI generates a short assessment from the development artefacts, and team members answer independently. The score maps to Naur's three layers of program understanding — domain intent, design justification, and modification capacity.
I'd love to hear from anyone who has tried measuring comprehension in their teams, formally or informally.
Top comments (0)