Can GPT Detect Gaming the System in Text Replays?

Abstract

Detecting gaming the system in an educational setting is essential because such behavior prevents educators from accurately assessing students’ knowledge, which negatively impacts their learning outcomes. This study examines the potential of utilizing Large Language Models (such as GPT 4 and GPT 3.5) combined with prompt engineering to identify gaming behaviors within data from a digital system platform. Various prompting strategies are analyzed, and experimental results are compared to previous approaches.