Unveiling Uncertainty: Supporting Learners Through NLP-Driven Confusion Identification

Online learning has increased significantly in popularity over the past several years, driven by global events such as the pandemic and the accessibility offered by educational platforms such as Moodle, Brightspace and so on. However, online learning platforms present challenges, including limited access to support and a sense of disconnection among students. This research works to mitigate these challenges by identifying confusion in learners in online learning platforms by analyzing their posts in course discussion forums. We utilized the Stanford MOOCPosts dataset, evaluated the performance of various ma-chine learning (ML) models, and explored the effectiveness of a custom classification embedding model on the Cohere. This Artificial Intelligence (AI) platform provides access to Large Language Models (LLM) and natural language processing (NLP) tools through an application programming interface (API). Our findings highlight the utility of AI platforms and LLMs in identifying and classifying confusion in online learners. With a substantial potential for the classification task to be dealt with by a custom model running on a third-party platform, researchers can focus on developing conversational agents to support learners with their confusion in courses in online learning platforms.