The vexing problem of validity and the future of second language assessment

Abstract

Construct validity and building validity arguments are some of the main challenges facing the language assessment community. The notion of construct validity and validity arguments arose from research in psychological assessment and developed into the gold standard of validation/validity research in language assessment. At a theoretical level, construct validity and validity arguments conflate the scientific reasoning in assessment and policy matters of ethics. Thus, a test validator is expected to simultaneously serve the role of conducting scientific research and examining the consequential basis of assessments. I contend that validity investigations should be decoupled from the ethical and social aspects of assessment. In addition, the near-exclusive focus of empirical construct validity research on cognitive processing has not resulted in sufficient accuracy and replicability in predicting test takers’ performance in real language use domains. Accordingly, I underscore the significance of prediction in validation, in contrast to explanation, and propose that the question to ask might not so much be about what a test measures as what type of methods and tools can better generate language use profiles. Finally, I suggest that interdisciplinary alliances with cognitive and computational neuroscience and artificial intelligence (AI) fields should be forged to meet the demands of language assessment in the 21st century.