ABSTRACT
The aim of this chapter is to introduce a measurement model that directly relates item characteristics to qualitative aspects of performance, with the explicit goal of providing instructionally useful diagnostic information about an examinee. Traditional psychometric models, including in particular logistic item response theory (IRT) models (Lord, 1980; Lord & Novick, 1969), are used to represent individuals on a single continuum of proficiency, an assumption clearly at odds with conceptions of performance derived from recent developments in cognitive psychology and the field of artificial intelligence. While traditional psychometrics may be sufficient and even desirable for some testing decisions, they provide little information relevant to the diagnosis of examinee knowledge or processing skill (Frederiksen, 1986; Glaser, 1972). Quantitative scores are especially misleading when differences in performance are not adequately characterized as being sequentially ordered.
