There's some really smart thinking and sharp analysis in this paper (18 page PDF). Here's the gist (and I'm eliding a lot of the detail): we create learner profiles (LP) in order to make predictions about their learning needs, but these models are based on assessors observations and intuitions, which may be misled by the complexity and multi-dimensionality of actual learner models. Hence, for example, we might see repeated attempts to solve a problem and conclude that a learner is engaged, but a deeper machine learning (ML) analysis may read the same data as an indication that the student is not engaged. Hence, there is a need for validation of these models, and this paper proposes a framework establishing "a bidirectional mapping between stakeholder requirements for LP analysis and ML-driven learner profiles."
Today: Total: [Direct link] [Share]
