Computerized Adaptive Screening Test (CAST): An Examination of Test Validity and Test Fairness

摘要

The Computerized Adaptive Screening Test (CAST) is used by Army recruiters to predict prospective applicants' (i.e., prospects') performance on the Armed Forces Qualification Test (AFQT). CAST performance data were collected from 60 recruiting stations across the country throughout calendar year 1985. These data were matched to applicant tapes from Military Entrance Processing Stations (MEPS) to obtain AFQT scores and relevant demographic information. Data analyses indicate that CAST is quite good at predicting AFQT scores for the entire sample and for examinees grouped by sex and race (black or white). When corrected for restriction in range, the cross-validated validity estimate based on the whole sample is .86. Race and sex differences in prediction exist, but these differences are minor and they correspond to those differences found with most cognitive ability tests. CAST's accuracy at predicting subsequent classification into important AFQT categories (i.e., 1-3A and 1-3B) is also discussed.