learner-model

A guess at how a particular student learns, written down precisely enough that a teacher — human or machine — can plan lessons against it.

In machine teaching, the learner-model is the load-bearing part: the computed "best lesson" is only best for the modeled student. A model that leaves out the student's limits — small working memory, slow eyes, few examples — recommends lessons the real student cannot take (two-windows). The rule the evidence gives: the proxy predicts the human only as far as it shares the human's constraints.