How it works

On paper is the school's published sexual violence policy, graded by AI against 17 criteria under a strict rubric, with every point backed by a quote from the policy itself.

In practice comes from ratings: a short multiple-choice questionnaire about the reporting process. It shows only once a school has at least 5 ratings.

Both use the same letters:

A

80 to 100

B

70 to 79

C

60 to 69

D

50 to 59

F

below 50

Built from the Students for Consent Culture minimum standards, plus a few Onus additions that make a policy usable. Each criterion scores 0 (not addressed), 1 (addressed but vague, permissive or discretionary, like “may” or “where possible”) or 2 (an explicit, specific and enforceable commitment: binding words like “will” or “must”, with names, steps, numbers or dates a student could hold the school to). A category's score is its points over the points possible, times 100; the On paper score is the average of the 5 categories, from 0 to 100.

When a school publishes its procedures as a separate document, Onus grades the policy and the procedures together as one text, and every quote shows which document and section it came from.

The AI has to quote the exact clause behind every score. Code then looks for that quote, word for word, in the school's policy text. If it isn't there, the point doesn't count.

The gap is On paper minus In practice, in grade points.

On the map, each dot is coloured by its On paper grade, from green for an A to red for an F. A ring appears around a dot once the school has at least 5 real ratings: Onus ratings and public records, never sample ratings. A thicker ring means a bigger gap; a ring touching the dot means students rate the school worse than its policy, and a ring with a space before it means better. No ring means there aren't enough real ratings yet. Each school's panel shows On paper, In practice and the gap.

• 12.5 or less either wayAligned
• More than 12.5 to 37.5Some gap
• More than 37.5Big gap
• In practice beats On paper by more than 12.5Better in practice
• No public policy foundNo public policy

BC law says a school “must review its sexual misconduct policy (a) at least once every 3 years” (Sexual Violence and Misconduct Policy Act, s. 3 (1)), and must consult students when it does (s. 4).

For each school, Onus reads the effective, approved or last-revised date printed in its published policy (or, if the policy prints none, its procedures) and shows the exact line it came from. The next review date is that date plus three years. When a document prints no date, the panel says “No date published.”

This is about the published policy only. A school may have reviewed its policy and decided not to change it, or changed it without republishing a new date, so a passed date doesn't mean a school broke the law. It means the published policy is older than the review cycle.

The 2025 Sexual Violence Policy Act will replace this law, but it is not yet in force; until it is, the 2016 Act's three-year review applies.

Each school's panel counts its ratings by source, so you always know what a grade rests on.

Onus (2 so far). Real ratings from people who signed in with a school email, including judges during the event.

Sample (433). Seed ratings so every school shows an In practice grade before real ones arrive. They are not real people. Every school's sample centres on the same neutral middle, and only a school's own published numbers move it: where its annual report gives formal reports received and how many were investigated, that share sets how often a sample rater says the school took action. They will be deleted as real ratings come in.

Public records (16). Numbers from schools' own annual reports, each linked to the report. They are shown in the panel; they don't feed the grade directly.

Your rating is stored with no link to your account, dates are rounded to the week, and a school's results appear only after 5 ratings. Each rating has a private code that can delete it; we keep only a scrambled copy, and the code itself stays in your browser, never with your account. There are no free-text boxes in the questionnaire, so no names or stories can be stored. Read the privacy policy.

Gemini (gemini-3.5-flash, Google) graded 7 schools. Its free daily quota ran out after 7 of the 24 policies on October 4, because retries during a Google outage counted against it, so the other 18, JIBC, KPU, Langara, NIC, NLC, NVIT, OC, RRU, Selkirk, SFU, TRU, UBCO, UBC, UFV, UNBC, UVic, VCC and VIU, were graded by Claude (claude-opus-5-5, Anthropic) with the same rubric and the same word-for-word quote check. Each school's panel names its grader.

As a check, Claude also graded one of Gemini's schools on its own, from the policy text alone, before seeing Gemini's answers. That is a small sample: the check was planned for 5 schools, but 4 of them ended up graded by Claude. The two agreed exactly on 16 of 17 criterion scores (94%) and were within one point on all 17. BCIT: Gemini 36, Claude 39.

AI can be wrong. The quote check stops it from inventing policy text, but it can still misjudge how strong a clause is. Every score shows its quote so you can judge for yourself, and the second auditor above shows how often two models read the same policy the same way.

In practice is mostly sample data for now, so the map colours each school by its On paper grade and only draws a gap ring once a school has 5 real ratings.

Every number, quote and policy on Onus links to where it came from. See all sources.