How do we attest an early-warning / dropout-prediction model is fair across cohorts?
You attest fairness by measuring the model's predictions and error rates across the relevant cohorts on a fixed probe set, documenting any disparity, and showing ongoing monitoring for drift — a reproducible, comparable record rather than a one-time claim.…
Answer.
You attest fairness by measuring the model's predictions and error rates across the relevant cohorts on a fixed probe set, documenting any disparity, and showing ongoing monitoring for drift — a reproducible, comparable record rather than a one-time claim. Planisphere generates that cross-cohort behaviour-and-drift evidence as a sha-pinned record; it surfaces disparity and change for your review and mitigation, but "fair" is your policy and legal judgment, not a Planisphere certification.
The mark behind the answer.
ED Office for Civil Rights · discriminatory-AI nondiscrimination resource.
More on OCR AI Guidance.
Prepare evidence for OCR AI Guidance review.
First evidence record within 21 days of access · re-runs in a single business day. Planisphere measures model behaviour and emits a reproducible, sha-pinned record — it does not certify, file, or give legal advice.