Explainability to improve AI-supported decision-making in the educational domain

• Applicants: Dr. Maximilian Förster; Professor Dr. Mathias Klier
• Subject Area: Operations Management and Computer Science for Business Administration
• Term: since 2026
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 577699540

The emerging research field of Explainable Artificial Intelligence (XAI) addresses the opacity of AI models. Applied to recommender systems, XAI methods make AI-based recommendations understandable by automatically generating explanations, while not impairing the AI models’ performance. Conventionally, explainable recommender systems are primarily used to improve decision efficiency, such as on e-commerce platforms, with explanations being used to increase adoption and acceptance of recommendations. But decisions with significant consequences require more reflection, i.e. purposeful and critical analysis of information, knowledge, and experiences, aimed at achieving a deeper understanding to inform (future) decisions. Our proposed research project explores how XAI can improve reflection in AI-augmented decision-making, building on the transformational idea that AI can serve as a reflection partner. The underlying problem context is education, where explainable recommender systems have the potential to support users in making reflective educational path decisions. In this context, explainability can potentially help users understand why the AI recommends certain alternatives, critically analysing these recommendations, and gain deeper understanding on which alternatives fit their interests. In our proposed research project “Explainability to improve AI-supported decision-making in the educational domain”, we seek to answer the following research question: How can XAI methods used in recommender systems support users in reflective decision-making regarding their educational path? Within the field of Information Systems (IS), this research project aims at developing methodological foundations for XAI methods used in AI-based recommender systems to support users in reflective decision-making, and at empirically evaluating the behavioural consequences of using such XAI methods. The project comprises two work packages WP1 and WP2. In WP1, we develop a novel XAI method to automatically generate explanations alongside educational path recommendations with the aim to support users in reflective decision-making. The XAI method builds on three core elements, which are subsequently developed in separate tasks. In WP2, we empirically investigate the XAI method by means of a functional evaluation, online experiments, and a field experiment. WP1 and WP2 are closely linked in an iterative process in the form of multiple design cycles. The outcomes of these interlinked work packages will include methodological foundations for XAI methods that can improve reflection in AI-augmented decision-making. Furthermore, we expect empirical insights from the evaluation into how the XAI method influences user behaviour. These insights offer an initial understanding of how explainability can realize the transformative vision of AI being a reflection partner as well as how explainable recommender systems can influence educational path decision-making.

• DFG Programme: Research Grants
• International Connection: Israel
• Cooperation Partner: Professor Dr. Lior Fink