Higher Education Assessment in the AI Era: From Preventing Academic Dishonesty to Designing Evidence of Thinking

Abstract

Higher education assessment, particularly within the social sciences and humanities, is currently confronting a crisis and profound challenges driven by the rapid evolution and proliferation of Generative Artificial Intelligence (GenAI). Traditional assessment frameworks have long operated under the assumption that a final written product serves as reliable evidence of a learner’s cognitive capacity and understanding. However, this foundational premise has been undermined by the ability of GenAI to generate comprehensive outputs within a remarkably short timeframe. Consequently, assessment practice has become trapped in a vicious cycle of prohibition, evasion, and detection. This paper argues that rather than focusing on the containment of AI, there is an imperative need to redesign assessments to accurately measure core competencies: academic thinking, selective agency, and accountability. To ensure the validity of evaluation, the paper proposes a paradigm shift from prioritizing written end-products to emphasizing the visible evidence of cognitive processes. On this basis, the paper introduces four assessment models adapted for the AI era, alongside a framework for academic integrity anchored in the transparent utilization of AI and student accountability.