KONX: A Dynamic Approach for Explainable AI in Higher Education

Abstract

In the era of data-driven decision-making, Higher Education Institutions (HEIs) can greatly benefit from the potential of explainable Artificial Intelligence (XAI) to provide transparent and interpretable insights. This paper presents the KONX (CONNECTS) approach, a methodology that leverages semantic web technologies to create a dynamic and comprehensive knowledge graph for advanced predictive models in academic advising. The KONX methodology focuses on harmonizing heterogeneous educational data sources, enabling seamless data querying and manipulation. By incorporating a feedback mechanism, the KONX approach remains adaptable to changes in the academic domain, continuously updating and maintaining its knowledge representation. To practically apply and evaluate the proposed methodology, a prototype was implemented and tested on an experimental case study concerning student outcomes prediction. The implemented prototype includes a graphical SPARQL generator interface to streamline the construction of SPARQL queries in an integrated way. In this way, this paper proposes both a comprehensive XAI methodology and a holistic technological infrastructure for applying the methodology in real-time scenarios. By bridging the gap between AI decision-making and human-comprehensible explanations, the KONX approach enhances transparency and user trust in AI-driven systems in the education sector.

Ähnliche Arbeiten