A Youmi Lab research tool

A Personalized Math Course-Taking Recommendation System for High School Students

Personalized Plan turns a fixed, one-size-fits-all math sequence into a plan tailored to each student. Grade by grade, it recommends the pathway that best supports their goals. Choose a model to begin.

The theory-driven model runs now. Answer seven short questions and you get a four-year plan. The data-driven model is in development.
The challenge

One sequence for everyone leaves learning on the table

Most U.S. high-school students follow the same fixed math sequence (Algebra I, Geometry, Algebra II, then advanced courses) regardless of their prior performance, motivation, or interests. That one-size-fits-all plan ignores meaningful variation in which pathway is actually best for a given student.

Personalized course planning has stayed largely heuristic, resting on individual teacher or counselor judgment. Personalized Plan reframes course-taking as a sequence of decisions that can be tailored, grade by grade, to maximize a student's final outcome. It brings the rigor of optimal dynamic treatment regimes from precision medicine into education.

The default plan
Algebra I Geometry Algebra II Advanced
The same four steps for every student, every year. There is no room for a student who is ready to accelerate, or one who would thrive on a different route.
A personalized plan
Algebra I Geo + Alg II Pre-calc AP Calculus
Personalized Plan recommends a route fit to the student's starting point, identity, and goals.
The Personalized Plan model

Two models, built two different ways

Both models recommend a math pathway toward one of two goals, stronger math achievement or a better chance of enrolling in college, at one of two horizons: a single Grade 9 decision or an adaptive Grade 9–12 plan. What sets them apart is how the recommendation is produced.

Theory-driven model

Rules you can read

It encodes a published expert framework as clear if-then decision rules. Every recommendation traces back to an explicit rule, so you can see exactly why a course was chosen, and the same answers always produce the same plan.

Available now.

Data-driven model

Learned from data

It estimates the best pathway directly from a large longitudinal study using causal machine learning (TMLE), so it can pick up patterns no one wrote down in advance. The restricted-data version is in disclosure review.

Preview on simulated data.

Try the theory-driven model Try the data-driven preview
n>