Our Promise, Shown
How Our Bias Audit Works
Every education product says “trust us.” We'd rather show you the mechanism. Before a lesson reaches your child, it passes a five-dimension bias audit — the same tests we teach your child to run on the world.
Same ruler, every subject.
If the same sentence produces different treatment depending only on who is named in it — that difference is the bias. In either direction.
The five tests
- 1
The swap test
Same action, different actor — does the label change?
We take a sentence, swap who it's about, and compare the words. If the same act is “determined” for one figure and “aggressive” for another, the framing failed — no matter which direction it leans.
- 2
Who's telling the story
Whose voices are present — and whose are missing?
A lesson told entirely by one narrator gets flagged, and the missing firsthand account is offered alongside it. We never remove the account that's there; we add the one that isn't.
- 3
Framing
Does the wording lead your child to a conclusion?
“City Spends $2M on Park” and “City Wastes $2M While Schools Crumble” report the same fact. When framing tilts, we show both framings and ask what each word choice makes the reader feel — we don't pick a side for your child.
- 4
Loaded language
Does each charged word actually meet its definition?
“Riot,” “militant,” “regime” — every charged word must earn its dictionary definition from the documented facts, and the same word must be available to every subject that fits it. Words applied to one group but not another for the same conduct get flagged.
- 5
Confidence
Does certainty match the evidence — in both directions?
Contested claims asserted as settled fact fail. So does the opposite: well-documented history hedged into “some people say.” Manufactured doubt is treated as seriously as overclaiming.
What happens when something fails
If our words failed
We rewrite our own framing before the lesson ships. The fix always raises the floor — adding the same context for everyone — never stripping context from anyone.
If a source is biased
We don't hide it — we teach it. A slanted historical source becomes the lesson: your child sees the framing, runs the swap test themselves, and owns the discovery. Source bias is curriculum, not contraband.
The rules that keep the audit honest
- Facts are untouchable. The audit only ever touches wording, framing, selection, and narration. A documented fact is never softened, hedged, or removed to protect anyone's conclusion — accuracy outranks comfort.
- Groups aren't actors. Claims of power must name something that can actually act — a company, an agency, a named official. “Five corporations control most American media” names actors; vague claims about an ethnicity or ancestry “controlling” things fail the test, whoever they target.
- Every statistic ships with the same context. What the number measures, who collected it, and what else explains it — the identical context package for every group's statistics. Raw numbers for one group and careful hedging for another is the failure mode, in either direction.
- The audit itself must be able to fail. We test the engine against a library of pre-registered cases with known answers — including cases where the correct verdict is “this is fine, don't flag it.” Over-flagging legitimate content because of who it mentions is itself a bias, and we test for it.
- Math is verified, not audited. Math problems are checked symbolically — the answer key is proven correct by computation. The bias audit applies to narrative subjects: reading, history, science, and writing.
That's the whole mechanism. No secret sauce, no “trust us.”
Start the Free Assessment