Lei Bao
The Ohio State University

Cognitive representations for educational measurement and modeling

A popular assumption underlying many existing assessment methods is that 
students possess internal models or have general abilities levels. Under 
this assumption, those doing assessment look for consistency between 
student responses on correlated items in a test in order to infer the 
existence and quality of those internal models or abilities. Inconsistent 
data may be rejected as noise. However, research in education, cognitive 
science, and neuroscience show that the cognitive response is a dynamic 
context dependent process. Results from a measurement instrument are 
dependent on the interactions between students' internal characteristics 
and the specific settings of the instrument. Inconsistent student behavior 
on correlated items not only represents a common phenomenon but provides 
important information about the state of the student's thinking. From this 
perspective, the inconsistency of student behavior becomes part of the 
signal rather than noise.