What must be considered in developing Artificial Intelligence (AI) policy in varying faculties at the University of St Andrews?
Head Researcher: Tabitha Stuart
Researchers: Annabel Butcher Burr, Camilla Hanley, and Greta Shope
Editors: Edwin Brattselius Thunfors and Grace Risucci
Editor-in-Chief: Edwin Brattselius Thunfors
Executive Summary
The advent of generative artificial intelligence (GenAI) has triggered a profound and unresolved disruption within academia. Its capacity to simulate human expression and analysis challenges foundational academic tenets of originality, critical thought, and authentic scholarship. Institutional responses have predominantly been to craft university-wide policies aimed at preserving academic integrity. However, these initial frameworks often adopt a generic, one-size-fits-all approach, primarily focusing on the mechanics of citation and the prohibition of outright plagiarism. This fails to confront a critical problem: the pedagogical utility, ethical contours, and practical implications of GenAI are inherently shaped by disciplinary epistemologies and assessment practices. A tool that fundamentally undermines learning in one context may be a legitimate professional instrument in another. Therefore, broad-brush policies risk being both practically irrelevant and intellectually prescriptive, leaving students and educators without the nuanced guidance necessary for responsible engagement.
This report advances a core thesis: the appropriateness of Artificial Intelligence (AI) use in academia is dictated by a discipline’s dominant assessment methodologies, which in turn must inform specific policy needs. To interrogate this premise, we employ two key analytical frameworks. First, we position disciplines along a spectrum defined by their primary assessment output: Calculation-Based (e.g., Mathematics, Physics), Essay-Based (e.g., History, Philosophy), and Languages (encompassing both Modern Foreign Languages and Computer Science). Each mode presents a distinct set of opportunities and vulnerabilities regarding AI integration. Second, we differentiate between Direct AI Use (the submission of AI-generated output as one’s own work) and Indirect AI Use (leveraging AI as a scaffold for explanation, brainstorming, and/or editing). The ethical and pedagogical valence of these uses varies dramatically across the assessment spectrum.
The aim of this project is to move beyond abstract debate and provide a granular, evidence-based analysis of these disciplinary divergences. Through a mixed-methods study at the University of St Andrews, combining policy analysis with comprehensive student and staff surveys, we investigate real-world practices, attitudes, and policy gaps. This report presents a comparative analysis across four deliberate case studies, each selected to exemplify a point on our analytical spectrum.
The report is structured as follows. We first outline our Methodological Framework. This is followed by four deep-dive Disciplinary Case Studies: Humanities (representing essay-based assessment), STEM (centred on calculation and problem-solving), Computer Science (as a formal language and applied field), and Modern Foreign Languages. Each case study integrates a focused literature review with our empirical survey data to evaluate discipline-specific hypotheses.