Bias-Free Computerized Testing.

Summarized in this report is research from a project designed to investigate the utility of item characteristic curve theory and computerized adaptive testing as means of measuring and reducing ethnic bias and unfairness in ability tests. During the theory development phase, an item characteristic curve theory model of bias was developed and used in computer simulation studies which investigated the bias reduction and fairness properties of computerized adaptive testing. In addition, a methodology for detecting test item bias was developed and validated. In the application phase the bias detection methodology was applied to six sets of real test data. In addition, the bias-reduction properties of computerized adaptive testing were examined in a live-testing study conducted in a racially mixed high school. The results of this research indicate that item characteristic curve theory provides a viable model for detecting item bias the incidence of item bias in existing tests is small, but because of its potential adverse effects, ability tests should be carefully examined for possible bias black students have different psychological reactions to the conditions of testing than White students and computerized adaptive testing can improve ability measurement for black students.