OS TESTES SITUACIONAIS DE INTELIGÊNCIA EMOCIONAL COMO UM TESTE ADAPTATIVO COMPUTADORIZADO

Resumo

The present study aimed to assess whether an emotional understanding test (STEU) and an emotional management test (STEM) could benefit from being administered as computer-adaptive tests (CATs) without impacting the validity of the test scores. To this end, 11 item selection algorithms (ISAs) were benchmarked for their bias and efficiency. Two simulation studies were run using the same response patterns from the 688 participants used in the original validation study, the same 11 ISAs, but differed in their stopping rules (SRs). For the first study, one simulation was run for each ISA with the SR being standard error lower than  (ΔSE < ), the most commonly used stopping rule criterion. For the second study, k simulations were run for each ISA, for each test, with the SR being a fixed number of items between 1 and k, where k was the total number of items of the relevant test (32 for the STEU and 30 for the STEM). Results of the first simulation showed that testing with all ISAs resulted in accurate ability estimates, all of them having r > .98 between their estimates and the estimates calculated with the entire test. The first simulation also showed that, using the best performing ISA, 368 (53.5%) participants needed to answer at least one fewer item without loss of validity. The second study showed that the STEU stood to benefit the most, with the mean standard error (MSE) being minimized six items before the end of the test, though ISAs based on the Kullback–Leibler information performed worse. However, these ISAs also displayed slightly less bias, , than the Fisher information-based ones . For the STEM, no ISA minimized MSE levels before the end of the test, but up to six fewer items for the STEM and 15 fewer items for the STEU could be administered with a slightly higher tolerance of ΔSE < . These results indicate that the use of CAT methodology to administer these tests is viable, and EI testing stands to gain from using CAT tests. Future studies should test ISA performance with additional testing constraints.