Validate High Stakes Inferences by Designing Good Experiments, Not Audit Items: A Comment on 'Self-Monitoring Assessments Educational Accountability Systems'

摘要

The use of large-scale assessments for making high stakes inferences about students and the schools in which they are situated is premised on the assumption that tests are sensitive to good instruction. An increase in the quality of classroom instruction should cause, on the average, an increase in test scores. In work with a number of colleagues over the past decade, Dan Koretz has argued that the validity of high stakes inferences from school-level improvements in test performance can be threatened by strategic instructional activities that serve to "inflate" test scores (Koretz & Beguin, 2010; Koretz & Hamilton, 2006; Koretz, McCaffrey, & Hamilton, 2001).