Why we built OutFigure.
AI tutors are very good at getting the next answer right. OutFigure checks whether the learner can still do it a week later, without help.
Measured from outside the tutor.
A tutor’s own numbers are about the tutor: engagement, turns, whether the next answer was right. They’re useful, but they can’t tell you what lasted. OutFigure connects what the tutor did to the learner’s own later work, and says how far that evidence goes.
That’s why a receipt keeps “how far it reaches” and “how sure of the cause” apart, shows the evidence that didn’t count and why, and is never edited after it’s written. Teaching styles are compared in fair, random tests, and nothing is estimated until every check is in.
Where it came from.
OutFigure grew out of Cognivia, a research project on what learning interactions can tell you about what someone actually knows. Its engine was built to separate what a person did with help from what they did on their own, which matters for any tutor that helps.
The method comes from that research, but the research isn’t proof that the method works. Results from OutFigure’s customers will be published once they exist, and only if the customers agree.
Built by Manik Maurya.
Get in touch.
For a pilot on the logs you already have, or questions about the method.
- Enquiry form
- Send an enquiry
- hello@outfigure.in
Please don’t send learner data or anyone else’s personal data.