RETHINKING EXPLAINABILITY IN EDUCATIONAL ARTIFICIAL INTELLIGENCE: A CRITICAL SYSTEMATIC REVIEW OF MODELS, APPLICATIONS, AND ETHICAL DIMENSIONS

Abstract

The increasing integration of Artificial Intelligence (AI) into educational technologies has elevated the importance of explainability to ensure transparency, pedagogical relevance, and ethical accountability. This systematic review critically examines empirical studies on Explainable Artificial Intelligence (XAI) in educational contexts published between 2015 and 2025. Guided by the PRISMA 2020 protocol, the review synthesizes findings from 106 peer-reviewed studies across K–12, higher education, and teacher training environments. The analysis identifies six interrelated themes: (1) types of XAI models and techniques, (2) domains of application in educational technologies, (3) operationalization of explainability, (4) impacts on learning and teaching, (5) challenges and limitations, and (6) ethical and epistemological considerations. Results reveal that while technical implementations of XAI are expanding, their pedagogical grounding remains limited, with explanations frequently framed as system outputs rather than tools for epistemic engagement. Explainability is often operationalized through vague definitions and evaluated using affective or behavioral proxies, with minimal attention to learning outcomes or cognitive development. Furthermore, ethical and cultural considerations are inconsistently addressed, raising concerns about inclusion, equity, and user agency. The review proposes a tripartite framework—pedagogical, epistemic, and ethical explainability—to guide future research and design. It calls for participatory, theory-informed, and context-sensitive approaches to ensure that educational XAI fosters critical thinking, learner autonomy, and instructional equity.

Ähnliche Arbeiten