Knowledge State Networks for Effective Skill Assessment in Atomic Learning
This enables efficient personalized learning for fine-grained skill ontologies without lengthy assessments, benefiting online education platforms.
The paper tackles the problem of lengthy skill assessments in personalized online learning by introducing knowledge state networks, which predict a learner's full knowledge state from partial information, resulting in a significant reduction in the number of assessment questions needed.
The goal of this paper is to introduce a new framework for fast and effective knowledge state assessments in the context of personalized, skill-based online learning. We use knowledge state networks - specific neural networks trained on assessment data of previous learners - to predict the full knowledge state of other learners from only partial information about their skills. In combination with a matching assessment strategy for asking discriminative questions we demonstrate that our approach leads to a significant speed-up of the assessment process - in terms of the necessary number of assessment questions - in comparison to standard assessment designs. In practice, the presented methods enable personalized, skill-based online learning also for skill ontologies of very fine granularity without deteriorating the associated learning experience by a lengthy assessment process.