Toward adaptive explainable artificial intelligence : a systematic literature review across human, task context, and model dimensions

Abstract

This thesis presents the new and growing area of adaptive Explainable Artificial Intelligence (XAI) through a Systematic Literature Review (SLR) carried out under the PRISMA 2020 guidelines. The research investigates the possibility of explanations being adjusted dynamically in three core aspects: Human, Task, and Model. The review of the 43 studies (2015-2025) points out that although the three dimensions have been extensively studied taken separately, very little research has been done to unify them in an adaptive framework. Most of the existing methods are not yet focused on three dimensions at the same time. The results expose critical research areas that need to be investigated, among them, the absence of three-dimensional adaptation mechanisms and the standardization of evaluation criteria. To tackle these challenges, the thesis proposes a conceptual model for the Adaptive XAI, which underlines the necessity for dynamic and balanced explanations that can vary according to the human, the text context, and the model focus.