Computer-Adaptive Testing (CAT) was first created to generate a total test scaled score (or estimate of ability) as efficiently as possible – that is, with as few items as possible administered in a very short time. Each item selected to be administered to a test taker is selected by the computer based on the test taker’s response (successful or not) to the previous question. Through this process, the system can zero in on the person’s ability level fairly quickly. Commonly used CAT models are not designed to be diagnostic – e.g., to identify specific knowledge or skills individual students need to work on. Lacking a good understanding of what is under the CAT hood, some consumers believe (and are led to believe) that the testing approach does more than it really can. Models that could be more informative would generally require access to much larger item banks than most and considerably longer testing times.
To access the 2012 paper, “The Promise of Computer-Adaptive Testing — The Quest for Information” by Stuart Kahl, see below.
1-CAT