CHALLENGES OF DESIGNING AND ADMINISTERING COMPUTER-ADAPTIVE TESTS
Computer adaptive testing has had a long history in research and practical application worldwide. The large number of both open-access and commercial software platforms developed specifically for creating and administering computer-adaptive tests (CATs) evidences their growing popularity. In foreign language teaching and learning computer-adaptive tests (CАТs) allow getting most accurate results when measuring the level of language proficiency in the least time-consuming way due to selecting items that are most appropriate for each testee: not too easy or too difficult. In a CAT each further item is presented depending on the previous correct or incorrect response: if the answer is correct the next item will be more difficult; conversely, if the answer is incorrect, the next item will belong to a lower level. Language CATs can serve different purposes: placement, achievement, proficiency, diagnostic testing. However, test designers face a number of challenges related to creating valid, reliable, objective and quick CATs: adaptive tests should be based on a large bank of finely-calibrated items; there should be instruments to evaluate the validity of test items. There are a number of more technical issues that need to be tackled: which item or level should be the starting point of a test, at what point can a test end, if it can be finished on student’s initiative, how many questions a student should answer before the test provides them with a higher/lower level items, how many questions a test should contain in total to present the most valid results, what item types allow for more objective measurement and can be properly delivered, how large the item bank should be. The designers of the few existing foreign language adaptive tests have approached these problems differently and despite their generally high reliability, these CATs have their limitations. Therefore, any effort put into designing and studying language CATs would make an important contribution.