Validity from the Perspective of Model-Based Reasoning

From a contemporary perspective on cognition, the between-persons variables in traitbased arguments in educational assessment are absurd over-simplifications. Yet, for a wide range of applications, they work. Rather than seeing such variables as independently-existing characteristics of people, we can view them as summaries of patterns in situated behaviors that could be understood at the finer grainsize of sociocognitive analyses. When done well, inference through coarser educational and psychological measurement models suits decisions and actions routinely encountered in school and work, yet is consistent with what we are learning about how people learn, act, and interact. An essential element of test validity is whether, in a given application, using a given model provides a sound basis for organizing observations and guiding actions in the situations for which it is intended. This presentation discusses the use of educational measurement models such as those of item response theory and cognitive diagnosis from the perspective of model-based reasoning, with a focus on validity.