Operationalizing a Human-Centered Explainable AI Framework for Adaptive, Trust-Aware Learning Analytics

Incorporating Artificial Intelligence in Learning Analytics, it is being increasingly utilized to forecast student performance, pinpoint students at risk and to aid educational decision making. Explainable Artificial Intelligence can help create more transparent AI generated predictions by providing reasons for the prediction. This explanation is not necessarily appropriate for every educational stakeholder as they all have different roles, levels of trust in AI and information-processing needs. This work further extends an existing proposed Human-Centered Explainable AI (HXAI) framework and operationalizes the main constructs and an adaptive mechanism for providing explanations in Learning Analytics. We introduce explainability, cognitive load, trust, educational role and effectiveness of decision support into the proposed framework. An Adaptive Explanation Engine takes these factors into consideration when deciding on an appropriate Explanation Delivery Mode (Summary, Moderate, or Detailed). The paper describes the operation of each construct, presents initial adaptation guidelines, and summarizes a DSR-based progression toward prototype development and subsequent empirical evaluation. The study offers an implementation-driven approach to adaptive explanation delivery in Learning Analytics that follows the principles of Human-Centered XAI.