Beyond the bluebook: a calibrated framework for redesigning construct-valid assessment in higher education in the age of generative AI

The rapid proliferation of generative artificial intelligence (GenAI) has complicated the link between what a student submits as evidence of learning and what that student actually knows and can do. In response, this study advances an AI-resilient approach to assessment design through the development of a calibrated framework for redesigning construct-valid assessment, based on a systematic review and synthesis of assessment frameworks and structured approaches. Following PRISMA 2020 reporting guidance, sixteen primary assessment frameworks and structured approaches were identified. The synthesis brought their complementary components together into three interdependent design constructs: authorisation, verification, and accountability. Messick’s unified validity theory provides the conceptual anchor connecting these components to the inference an assessment is intended to support. The synthesis revealed that the central gap in the literature is not a lack of relevant approaches or components, but calibration. This gap informed the development of the Calibrated Framework for AI-Resilient Assessment (AIRA), which proposes that verification and accountability should become stronger and more diverse as authorised GenAI involvement increases or becomes more consequential to the intended construct. The framework provides educators with a guide for designing assessments and evaluating student learning in AI-integrated environments.