That is not a marketing promise. It is a constraint written into the instructions the AI receives before every debate, and you can read what it says below.
Students practice disagreeing well: reconstructing a position they do not hold, asking why rather than what, and ending a hard conversation with the relationship intact. They are scored on those specific behaviors, not on which side they took.
Identify what the disagreement is actually about — the underlying value tension, not a "some people think X" summary.
Present each side at its strongest. If you cannot articulate why a smart, moral person would hold a position, you have not understood it yet.
The AI is instructed not to tell a student which side is right. If it catches itself concluding with a recommendation on a contested question, it must replace that with a question that returns the judgment to the student.
If the way a scenario was presented favors one position through word choice, ordering, or emotional weight, the AI must say so.
The protocol applies to value disagreements and policy tradeoffs. It does not apply to empirical questions with scientific consensus — climate change, vaccine safety, and the historicity of the Holocaust are named in the code as settled, not contested.
Students aged 13 to 17 trigger a parental consent step at signup: a consent link is sent to a parent address, and the account is not activated until the parent confirms it. Signups under 13 are deleted. Age is self-reported at signup, so this step protects students who give their real date of birth; a school that needs verified consent should collect it through its usual process.
Teachers create classes and see their own students' progress. Facilitator and moderation views exist for supervised sessions. The free drill at /try requires no account at all, so you can evaluate the experience before asking anyone to create one.
Before adopting this platform, ask us for whatever your district requires — a data processing agreement, a privacy review, an accessibility statement. If we do not yet have a document you need, we will tell you that plainly rather than send you something that does not apply.
The literature supports the premise that empathy is trainable, at a moderate effect size. It does not yet show that this specific platform changes anything — that study is drafted and has not been run.
We would rather you hear that from us than discover it in a procurement review.