Validity at Risk? the Future of Generative AI Use by Candidates in Assessments

ABSTRACT In this commentary on Lievens and Dunlop (2024), we critically evaluate the implications of candidate GenAI use from the perspectives of criterion‐ and construct‐related validity. While the focal article provides valuable scenarios for understanding the role of GenAI use by candidates, we argue it may underestimate the speed of technological development and overstate the generalizability of augmentation across all job types. By considering increasing GenAI accessibility, we highlight how ease of use may weaken the diagnostic value of GenAI‐related variance in assessments. Furthermore, we propose experimental paradigms to empirically investigate the impact of GenAI on criterion‐ and construct‐related validity by framing GenAI as either an impersonator or an assistant. Finally, we reflect on a future where GenAI agents may challenge existing conceptions of construct validity, calling for a rethinking of the way we define and measure psychological constructs in the age of AI.