Next-generation Enem assessment with fewer items and high reliability using CAT

Alexandre Jaloto; Ricardo Primi

doi:10.1590/SciELOPreprints.5339

##article.authors##

Alexandre Jaloto Instituto Nacional de Estudos e Pesquisas Educacionais Anísio Teixeira https://orcid.org/0000-0002-5291-1768
Ricardo Primi Universidade São Francisco

DOI:

https://doi.org/10.1590/SciELOPreprints.5339

Keywords:

student assessment, psychometrics, item response theory, admission to higher education

Abstract

The Brazilian High School Exam (Enem) results are used to enter higher education and obtain scholarships and study financing. The Enem comprises an essay and four 45-item tests that measure educational skills in different areas of knowledge. Therefore, this work aims to verify the possibility of reducing the number of items in Enem through a Computerized Adaptive Testing (CAT). We used tests from the 2009 to 2019 editions of Enem. The items have been calibrated and positioned on four scales (one for each area). we simulated a CAT, which ended when the error was less than or equal to 0.30, or 45 items were applied. On average, the application ranged from 12.0 (Languages and Codes; LC) to 29.2 (Mathematics; MT) items. Results point to the potential of reducing the size of Enem to 20 items for a proportion that varies from 39.8% (MT) to 94.8% (LC).