Rethinking University Assessment in the Age of LLMs: Practices, Tools and New Literacies

Abstract

The widespread adoption of Large Language Models (LLMs) challenges traditional approaches to university assessment by making learning processes less visible and authorship more difficult to interpret. Drawing on an exploratory case study conducted within a General Didactics course, this paper examines students’ uses, perceptions, and interactions with generative AI, highlighting the limitations of product-oriented assessment. It proposes a process-based framework centred on transparency, verifiability, and attribution through AI usage agreements and logs, together with redesigned assessment tasks and criteria. The paper argues that assessment literacy must evolve to foster responsible AI use while preserving students’ agency, critical thinking, and reflective learning.