A systematic review of explainability approaches in designing AI interfaces and analyzing the role of visual feedback in shaping user trust.
A systematic review of explainability approaches in designing AI interfaces and analyzing the role of visual feedback in shaping user trust.
Ebadollah Zohre Wendy1 sarahpourasadollahi2
1) Faculty member, Malayer University, Computer Department, Malayer Branch, Islamic Azad University, Malayer, Iran.
2) Master s student in Software Engineering, Islamic Azad University, Malayer Branch, Malayer, Iran.
IOI Code
:
AAHD-ABKKFC
Publication :
National Conference on Information Technology, Nanotechnology, Artificial Intelligence and Technological Futures Studies(itnaf.ir)
Abstract :
With the increasing penetration of artificial intelligence (AI) systems in sensitive and high-risk domains, the issue of the “black box” nature of these systems and the lack of transparency in their decision-making processes has become a major challenge. This lack of transparency is a major obstacle to the formation of effective and calibrated user trust. Explainable Artificial Intelligence (XAI) has emerged as an interdisciplinary field with the aim of making these processes transparent. However, simply producing an explanation is not enough; how this explanation is presented through the user interface (UI) plays a crucial role in its effectiveness. This article, using a systematic review method and qualitative-conceptual analysis, comprehensively examines explainability approaches in AI interface design and, in particular, analyzes the key role of visual feedback in shaping, calibrating, and maintaining user trust. By analyzing and reviewing the research literature published between 2017 and 2024, this study presents a taxonomy of XAI techniques and their corresponding visual feedback mechanisms. Findings show that visual feedback, by translating abstract data of explanations (such as feature importance or counterexamples) into understandable representations (such as heat maps, bar graphs, or natural language), acts as a cognitive bridge and directly influences the antecedents of trust (perceived competence, benevolence, and integrity of the system). The results suggest that no single visual approach is optimal for all scenarios and its effectiveness strongly depends on factors such as the type of AI model, task complexity, user expertise, and the purpose of the explanation (e.g., debugging or decision justification). This paper ultimately presents an integrated conceptual model that maps the dynamic interactions between XAI techniques, visual feedback design, cognitive antecedents of trust, and end-user behavior, and identifies future research directions toward designing human-centered and trustworthy AI interfaces.
Keywords : explainable artificial intelligence (xai) trust in ai user interface design human computer interaction (hci) visual feedback information visualization