Unconfound Labs Back to the studio
Learning Systems

An engine for investigative learning.

Unconfound Labs is building a system that drops learners into investigative environments on realistic, messy data: open-ended cases where the answer isn't clean, the records don't line up, and progress comes from reasoning under uncertainty the way analysts actually do. It applies the studio's own field method, causal inference and data linkage, at a level a classroom can run.

What students investigate

The cases we've built so far.

Each case is anchored in a real domain and run on realistic, simulated data, messy and incomplete but safe to share. A pilot can start with one and grow from there.

Public Health · Outbreak

The Holloway Creek Outbreak

Campers are getting sick on a staggered timeline, a few of the sick have no link to the camp at all, and the lab results are days away. Learners work the theories the way an epidemiologist would: narrow the source, weigh the evidence, and defend the call.

Flagship case
Sports · Performance Analytics

The Hot Hand

A GM is about to extend his star on a career contract, but the offense stalls every time the game is close, and the building has six theories about why. Learners clean a messy shot log and weigh what's really going on: nerve, fatigue, the defense, a regular-season number that may not hold up in the playoffs. The data won't hand over one clean answer, which is the point.

Ready to run
The shift
No answer key. A case, the evidence in front of you, and a call you have to defend.
Why it works

Closer to real analysis than a worksheet.

Realistic, not real

Cases run on simulated data engineered to behave like the real thing: messy, incomplete, and safe to put in front of students, with no sensitive records involved.

Reasoning over recall

There's nothing to memorize. Learners build a case, test it against the evidence, and have to defend it, the work that survives scrutiny in the field.

Built on field method

The same causal-inference and data-linkage rigor the Labs applies to public-health and pharma data, brought down to a level a classroom can run.