Thinking to learn: managing the risks of outsourcing to GenAI

Abstract

This conceptual paper addresses the pedagogical and ethical challenges posed by generative artificial intelligence (GenAI) in higher education assessment. While GenAI can enhance productivity and innovation, it introduces challenges related to cognitive offloading, academic integrity and assessment validity. Responding to sector-wide calls for structural reform, this paper proposes a developmental assessment framework that enables educators to make principled decisions about GenAI use. Drawing on recent scholarship, it synthesises key concerns, cognitive offloading, ethical boundaries, pre-existing assessment challenges and design frameworks, to establish the pedagogical and regulatory foundations for assessment transformation. The paper’s contribution is twofold: it applies Griffin’s developmental assessment model to GenAI contexts, and it combines that model with the AI Assessment Scale (AIAS) to operationalise principled decisions about where GenAI may ethically support assessment without compromising validity. The framework guides educators through aligning learning outcomes with assessment tasks, unpacking constructs, identifying capabilities and mapping observable indicators to AI use. This structured approach complements tools such as the AI Assessment Scale (AIAS) by embedding validity and cognitive ownership into assessment design. It supports transparent, defensible decision-making and shifts the focus from rule enforcement to thoughtful design. Implications for educators, students, institutions, and integrity and governance are explored, with attention to implementation and the need for sustained development. In doing so, this paper reframes assessment integrity as a design challenge rather than a compliance issue, contributing to a more principled and inclusive approach to learning in the GenAI era.