Title: How far do stray CATs wander? On the robustness of CAT to persons who initially do not respond in line with their true ability.
Abstract: In computerized adaptive testing (CAT), the idea is that with each item presented to and responded by you, the test administration algoritm can learn more about you and in turn can tailor the test towards your ability level with each subsequent item. This sequential learning-and-tailoring is the essential principle that allow CATs to still deliver accurate measurement but in a more efficient manner leading to short yet effective tests. We investigate here to what extent CAT is robust to people who initially do not provide responses in line with their actual ability level (e.g., misstart or warming up).