Predictive models built in a modelable ecosystem generalize across institutions¶

Claim · All claims
Evidence · 1 study (2 entries) · design r2 · q2
Currency · newest study 2026 · checked against its sources 2026-10-08 (automated, 2 of 2 entries) · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? A system designed with modelability in mind can produce predictive models that generalize effectively across diverse educational contexts. → Alice Xu 2026 q2 i? Model generalizability was tested by evaluating performance in a different course, institution, instructor, and course length from the development setting. → Alice Xu 2026 (2)

Evidence¶

Alice Xu 2026¶

Alice Xu, Icy (Yunyi) Zhang, Adam B. Blake and James W. Stigler. (2026). Modelability as a Strategy for Improving the Generalizability and Scalability of Predictive Models. Journal of Learning Analytics, 13(1). https://doi.org/10.18608/jla.2026.9099

q2 · i? · design · r2

The article's reported result for its cross-institutional case study: models developed on CourseKata data from one institution were tested on another, and the article reports they "generalize effectively across diverse educational contexts". No effect size or performance statistic is printed in the available text.

"Results show that a system designed with modelability in mind can produce predictive models that generalize effectively across diverse educational contexts."

Alice Xu 2026 (2)¶

Alice Xu, Icy (Yunyi) Zhang, Adam B. Blake and James W. Stigler. (2026). Modelability as a Strategy for Improving the Generalizability and Scalability of Predictive Models. Journal of Learning Analytics, 13(1). https://doi.org/10.18608/jla.2026.9099

q2 · i? · design · r2

Case study design: models were developed on one introductory statistics course and evaluated on a second course using the same CourseKata textbook, differing in instructor, institution, and course length (ten-week vs. accelerated six-week session).

"We then evaluated the model's generalizability in a different course using the same textbook taught by another instructor, at a different institution, and over a different course length."

Discussion¶

Related Claims¶

• Attendance, behavior, and course performance (the ABC indicators) reliably predict which secondary students are at risk of dropping out — the general claim this finding bears on, with the studies behind it
• Behavioral-only metrics from an instrumented online textbook support early prediction of final course grades in flipped statistics courses — related
• A predictive model built from school and child welfare data identifies students at risk of near-term academic problems — related
• A predictive model built on school and child welfare data was assessed for identifying at-risk students in two Allegheny County education agencies — related
• Current AI deployments in higher education remain task-oriented rather than ecosystem-oriented despite advances in LLMs and multimodal AI — related