I will use about 5 days to introduce the basic principle and approach of computerized adaptive test and its theoretical foundation, Item Response Theory (IRT).

If you’ve ever taken part in the new GRE test, you might notice that the test was not a set of sequential items, but divided into five independent parts (two for verbal, two for math, and one random part). Actually, when you did the first part of the test, the questions in the rest four parts were not decided. These questions would be chosen from the questions pool in accordance with your performance in the previous test parts. Generally, the better you behaved in the previous parts, the harder the questions you might meet in the next test part.

That’s exactly a kind of computerized adaptive test. During the test process, the algorithm inside the computer tried to constantly evaluate the ability of the tester. Compared with the traditional test method, the adaptive test shares some great advantages, the dynamic estimation made the test shorter but more valid and reliable.

It is highly recommended that we apply the adaptive test into our product and service when our questions database gained into a mature system. It is still a fangle, but makes easy to offer a personalized experience to our users.