VizChat: Enhancing Learning Analytics Dashboards with Contextualised Explanations Using Multimodal Generative AI Chatbots

Learning analytics dashboards (LADs) serve as pivotal tools in transforming complex learner data into actionable insights for educational stakeholders. Despite their potential, the effectiveness of LADs, particularly the visualisations they utilise, has been under scrutiny. Concerns have been raised about their potential to cause cognitive overload, especially for users with limited data visualisation literacy, thus questioning their practical utility in supporting decision-making and reflective practices. This tool paper tackles these concerns by introducing VizChat, an open-sourced, prototype chatbot designed to augment LADs by providing contextualised, AI-generated explanations for visualisations. Developed on multimodal generative AI (GPT-4V) and retrieval-augmented generation (Langchain), VizChat offers on-demand, contextually relevant explanations that aim to improve user comprehension without overwhelming them with excessive information. Through a case study, we demonstrated VizChat’s diverse capabilities, including actively seeking clarifications on ambiguous queries, personalising responses based on previous user interactions, providing contextually relevant explanations of specific visualisations, integrating information from multiple visualisations for a comprehensive response, and offering detailed insights into the data collection and analysis processes behind each visualisation. Such efforts support the paradigm shift from exploratory to explanatory approaches in LADs, highlighting the potential of integrating generative AI and chatbots to enhance the educational value of learning analytics.