Abstract

The rapid adoption of Artificial Intelligence (AI) in organizational decision-making has increased the need for transparent and trustworthy intelligent systems. Explainable Artificial Intelligence (XAI) has emerged as an effective approach to improving users' understanding of AI decision-making processes. However, recent studies suggest that providing more explanations does not necessarily enhance trust; rather, the effectiveness of explanations depends on their ability to support users' cognitive processing. This study aims to develop a conceptual framework explaining how XAI features influence calibrated trust through the mediating role of cognitive load. A systematic literature review (SLR) was conducted following the PRISMA 2020 guidelines. Relevant studies were retrieved from major international and Persian scientific databases, and after applying predefined inclusion and exclusion criteria, 33 studies (26 international and 7 Persian) were selected for thematic synthesis. The findings indicate that XAI features—including transparency, interpretability, completeness, simplicity, relevance, and explanation quality—improve users' perceived explanation quality and mental model formation. These mechanisms reduce extraneous cognitive load, facilitate meaningful cognitive processing, and promote calibrated trust rather than overtrust or distrust. Furthermore, the relationships among these constructs are moderated by individual and contextual factors such as AI literacy, user expertise, prior experience, task complexity, and explanation type. Based on the synthesized evidence, this study proposes an integrated conceptual model linking XAI features, cognitive load, and calibrated trust. The proposed framework contributes to the theoretical integration of explainable AI, cognitive load theory, and trust in automation while offering practical implications for designing human-centered AI systems in information technology management and intelligent business environments.