Question · OCR AI Guidance
How do we show an early-warning / dropout-prediction model's results across cohorts?
OCR's November 2024 resource gave 21 examples of how school AI use could lead to investigations under Title VI, Title IX and Section 504, such as biased facial recognition, AI proctoring that flagged a disability, and unreviewed AI-written 504 plans.
01 / The answer
The short answer
Compare the model's predictions and error rates for each student group that matters, such as race, sex, disability and English learner status, using the same data and time period. Report how often each group is flagged, how often flags are wrong in each direction, and how results change over time as the model or student population changes. Record what staff do with flags, because those decisions are what affect students. Repeat the check whenever the model, its inputs or its use changes.
02 / The rule
What OCR AI Guidance requires
OCR's November 2024 resource gave 21 examples of how school AI use could lead to investigations under Title VI, Title IX and Section 504, such as biased facial recognition, AI proctoring that flagged a disability, and unreviewed AI-written 504 plans. ED has since marked it formally rescinded, but the underlying statutes still apply, though Title VI enforcement now covers intentional discrimination only.
- 01Underlying law: do not treat students differently by race, color, national origin, sex or disability through AI tools, directly or through vendors.
- 02Underlying law: respond to known harassment, including AI-generated harassment such as deepfakes, under Title VI, Title IX and Section 504.
- 03Underlying law: provide academic adjustments when AI proctoring or monitoring misreads disability-related behavior.
- 04Underlying law: have knowledgeable people review AI-drafted Section 504 plans and IEPs for each student's individual needs.
03 / Records
Records to keep
- 01Vendor bias and accuracy reviews for AI tools affecting students
- 02Harassment complaint logs covering AI-generated content
- 03Accommodation requests and responses tied to AI proctoring
- 04Records of human review of AI-drafted 504 plans
Optional evidence · where Planisphere fits
Keep a record anyone can check
When your team meets a duty, Planisphere can seal the event it reports as a receipt.
A disclosure shown, a review signed, a model change approved: the event your system reports.
The event becomes a sealed receipt, dated and attributed to whoever is signed in.
Paste the receipt into Verify. No account or key.
A record supports your compliance evidence process. It is not a legal finding or a certification. Systems that call Planisphere's gate API can also hold or refuse an action before it runs.
SOURCES
Go back to the source
U.S. Department of Education, Office for Civil Rights (copy hosted by Connecticut DAS)
U.S. Department of Education
RELATED
Related: OCR AI Guidance: the full guide · What does ED OCR's AI guidance require schools to check for? · How do we show OCR we measured our school AI for discrimination?