Modelability: designing learning ecosystems to support generalizable predictive models via three principles¶

Theory · All theories
Evidence · 1 claim (1 for) · 1 study (1 design), q2 · 0 of 1 report an effect size · 1 claim rests on one study

Description¶

Modelability is defined as "the extent to which a system is designed to support the development of models that are generalizable". The article argues a highly modelable system includes three features: valid and interpretable measurements, scalable and stable implementation, and a collaborative research-practice-technology ecosystem. Rather than improving algorithms post hoc, the approach improves the affordances of the world being modeled, addressing generalizability at its source.

Design Implications¶

Context¶

Requirements¶

• Systems must generate valid and interpretable measurements, be implemented at scale with consistency, and be sustained through collaboration among researchers, practitioners, and developers.

Constraints¶

• Algorithmic improvements alone can only enhance generalizability within the constraints of the data itself; highly context-dependent training data limits any model.

Target Learners¶

• students in online and blended higher education and K-12 courses

Target Learning Objectives¶

• predicting learning outcomes such as course grades to enable early intervention

Claims¶

• Cross Institutional Generalizability Coursekata Models [+M]

Related Theories¶

• Research-practice-technology partnerships (RPTPs) as the ecosystem sustaining modelable systems

Examples¶

• Decide what stays constant and what varies, based on the research goal, before scaling across contexts
• CourseKata fully instrumented online textbook platform

Key Sources¶

• 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