MOOCs Meet Measurement Theory: A Topic-Modelling Approach

摘要: This paper adapts topic models to the psychometric testing of MOOC students based on their online forum postings. Measurement theory from education and psychology provides statistical for quantifying a person's attainment intangible attributes such as attitudes, abilities or intelligence. Such infer latent skill levels by relating them individuals' observed responses series items quiz questions. The set can be used measure if conform Guttman scale. well-scaled differentiate between individuals inferred span entire range most basic advanced. In practice, researchers manually devise (quiz questions) while optimising conformance. Due costly nature expert requirements this process, has found limited use in everyday teaching. We aim develop usable measurement highly-instrumented delivery platforms, using participation automatically-extracted topics items. challenge is formalise scale educational constraint incorporate it into models. To favour that automatically scale, we introduce novel regularisation non-negative matrix factorisation-based modelling. demonstrate suitability our approach with both quantitative experiments three Coursera MOOCs, qualitative survey interpretability two MOOCs domain interviews.