An attempt to cheat using GPT-4: findings, discussion and recommendations for academic staff and students

Abstract

This manuscript explores the use and detection of ChatGPT artificial intelligence (AI) content, in the academic setting.  The study involves faculty and student partnership at a Higher Education Provider (HEP), examining the implications of the chatbot's use in academic writing and its potential impact on educational practice.

We employ a dialogical approach in this Scholarship of Teaching and Learning (SoTL) study, mirroring the ChatGPT conversational style.  A former student uses ChatGPT-4 to rewrite an essay, and the generated content is evaluated by free-to-use AI checking tools, Turnitin, and the module tutor.

Our findings illustrate that while ChatGPT-4 could generate a passable assignment, it lacks depth, criticality, and contextualisation.  Further, AI checking tools, while useful, exhibit limitations. However, the generated content could serve as a starting point for students, suggesting a possible shift in the educational paradigm with AI's integration.

The manuscript also underscores the importance of understanding and navigating AI responsibly, calling for clear guidelines for AI use and improved methods of detecting AI generated content.  While AI has emerged as an efficient tool for generating academic content, our study highlights the necessity of human-centric learning and critical thinking skills in academia, emphasising that AI should not replace personal research.  We also discuss potential changes to assessment methods in response to AI advancements, recommending strategies for both academic staff and students.

Our findings have implications for educational institutions worldwide as AI tools become increasingly accessible, raising questions about academic integrity, pedagogical strategies, and the evolving relationship between technology and education.