Using contextual factors analysis to explain transfer of least common multiple skills


Transfer of learning to new or different contexts has always been a chief concern of education because unlike training for a specific job, education must establish skills without knowing exactly how those skills might be called upon. Research on transfer can be difficult, because it is often superficially unclear why transfer occurs or, more frequently, does not, in a particular paradigm. While initial results with Learning Factors Transfer (LiFT) analysis (a search procedure using Performance Factors Analysis, PFA) show that more predictive models can be built by paying attention to these transfer factors [1, 2], like proceeding models such as AFM (Additive Factors Model) [3], these models rely on a Q-matrix analysis that treats skills as discrete units at transfer. Because of this discrete treatment, the models are more parsimonious, but may lose resolution on aspects of component transfer. To improve understanding of this transfer, we develop new logistic regression model variants that predict learning differences as a function of the context of learning. One advantage of these models is that they allow us to disentangle learning of transferable knowledge from the actual transfer performance episodes. © 2011 Springer-Verlag Berlin Heidelberg.

Publication Title

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)