Comparison of human and machine-based educational standard assignment networks

Increasing availability of digital libraries of K-12 educational resources, coupled with an increased emphasis on standard-based teaching necessitates assignment of the standards to those resources. Since manual assignment is a laborious and ongoing task, machine-based standard assignment tools have been under development for some time. Unfortunately, data on the performance of these machine-based classifiers are mostly lacking. In this article, we explore network modeling and layout to gain insight into the differences between assignments made by catalogers and those by the well-known Content Assignment Tool (CAT) machine-based classifier. To build the standard assignment networks, we define standards to be linked if they are jointly assigned to a learning resource. Comparative analysis of the topology and layout of the networks shows that whereas the cataloger-based network reflects the underlying curriculum, i.e., clusters of standards separate along lines of lesson content and pedagogical principles, the machine-based network lacks these relationships. This shortcoming is partially traced back to the machine classifier’s difficulties in recognizing standards that express ways and means of conducting science.