This module reframed the failure. I had been reading a low score as data about the student. It is more often data about the design. Barriers live in the learning environment, not in the learner, and most of the ones in my classroom I built myself.
The clearest example is the confounded goal. When I assess procedural competence through a written task, I am testing two things at once, the skill and the mode, and I cannot tell which one produced the score. Separating the goal from the means is the fix. If the learning target is safe operation of the equipment, then the assessment has to require safe operation of the equipment, and nothing about the format should stand between a competent student and demonstrating that competence.
Applying it means auditing my assessments question by question and asking what each item is actually measuring. It means running the barrier question before the assessment goes out rather than after the scores come back, because once I am looking at a failed attempt I am rationalizing. It means giving students more than one route to show the same competency, and using frequent low-stakes checks so I am gathering evidence while the learning is still in progress instead of at the point where I can no longer act on it.
Parents and guardians are also a formative data source I have almost entirely ignored, and a start-of-year survey asking how a student learns and communicates at home is a low-cost way to get context I currently wait months to discover.
