A Human-centric Approach to Explain Evolving Data: A Case Study on Education

This study focuses on how artificial intelligence (AI) can be used in education while emphasizing the importance of adhering to European regulations requiring explanations of automatic methods. The study uses a prototype-based dynamic incremental classification algorithm called Dynamic Incremental Semi-Supervised Fuzzy C-Means - DISSFCM, based on Fuzzy C-Means (FCM), that leverages fuzzy logic to analyze educational data related to students' interactions with a learning platform. In this work, we propose employing fuzzy logic to generate humancentric explanations of the dynamic process in terms of IF-THEN rules derived from the DISSFCM prototypes obtained at different time frames. The Open University dataset (OLUD) is used for experimentation and validation. The study demonstrated that the dynamic algorithm could adapt its model based on changes in data from one chunk to the next. The explanation model proposed in the study was found to be effective in describing the evolving process using fuzzy terms that were easy for stakeholders to understand while also being robust to variations in the percentage of data labeling that may occur in real-world applications.