When Students Use AI: Toward Validity-Centered Assessment Design

The rapid normalization of generative artificial intelligence (GenAI) in higher education has intensified debates about cheating and academic integrity. Yet a growing body of assessment scholarship argues that the more fundamental issue is validity: whether assessment scores continue to support the interpretations and decisions that assessors and institutions want to make in contexts where students can legitimately or illegitimately rely on powerful GenAI tools. Drawing on argument-based validity theory, this paper re-articulates the interpretive argument for AI-inclusive assessments by distinguishing ten validity threats along an inference chain (domain, scoring, generalization, explanation, extrapolation, utilization). A main contribution of this paper is a Validity-Centered Assessment Design (V-CAD) checklist, an inference-by-inference diagnostic and evidence scaffold to help educators and assessment designers identify which inference is under threat, which mechanisms make it salient under AI inclusion, which design implication may follow, and which locally collectable evidence could strengthen or weaken the corresponding validity claim.