Large-scale testing in the face of AI

Abstract

Key digested messageThis article examines the expansive growth of ChatGPT and the implications for large-scale test design. The authors contend that the impressive test simulation results observed by Chat-GPT undergird ongoing construct validity concerns with student testing. In order to address these challenges, a set of strategies is proposed that emphasises authentic assessment, the importance of human elements in traditional paper-and-pencil questions, and the controversial issue of the stakes ascribed to test results. Collectively, these approaches are meant to help test developers more carefully consider existing limitations within traditional standardised and large-scale assessment programs. Ultimately, test design reforms that enhance validity are increasingly needed to address the challenges posed by AI applications.