B0818
Title: Cheaters' detection via response times in computerized adaptive testing
Authors: Luca Bungaro - University of Bologna (Italy) [presenting]
Bernard Veldkamp - University of Twente (Netherlands)
Mariagiulia Matteucci - University of Bologna (Italy)
Stefania Mignani - University of Bologna (Italy)
Abstract: An advantage of computerized-based testing is the possibility to collect, besides the item responses, the subjects' response times. In computerized adaptive testing, where the test is tailored to a single respondent, response times can be used to improve item selection and ability estimation and detect cheating. A method to identify subjects with item pre-knowledge by using response times is proposed. In particular, a new interim person fit statistics has been developed, which is able to identify cheaters during the test and, consequently, may lead to the change of the test itself by using a more secure item database and to the improvement of the ability and speed estimation. A simulation study shows the accuracy of the person fit statistics for correctly identifying the cheaters and improving the ability to estimate them under different scenarios.
