Online Maths for Social Science Assessment Design in the Age of Generative AI

In the fast-changing educational world, Generative AI (GenAI) has brought about big changes, especially in online formative and summative assessments. Universities are concerned about the unethical GenAI use, compromising academic integrity. This study proposes and evaluates strategies to design questions in maths for social science education that are challenging for ChatGPT-3.5 to solve. Drawing on Bloom’s Taxonomy, a trend analysis of academic performances and focus group discussions, it proposes a transformed approach to assessment design: the SHARP (Strategic, Holistic, Adaptive, Reflective, Process) assessment cycle. This framework is iterative and integrates real-time feedback to ensure inclusivity and transparency, stemming from a Reflect-Rewrite-Retest-Review redesign approach, focusing on higher-order cognitive questions. A quantitative analysis between 2020 and 2024 reveals a significant increase in higher-order level questions (e.g. from 29% to 84% in a test) and a significant but not drastic drop in academic performance. The effectiveness of ChatGPT-challenging designs is corroborated by focus group discussions, highlighting the need for a balance between student accessibility and academic rigour. This study contributes to the literature by providing unique empirical evidence on the validity of the strategies and offering actionable steps for educators, policymakers and institutions to maintain academic integrity in maths for social science education.