Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI)

This paper is a survey of explainable artificial intelligence (XAI), motivated by the fast and widespread adoption of AI and the key impediment that many AI-based systems lack transparency. Because their black-box nature yields powerful predictions that cannot be directly explained, the authors situate the new debate on XAI as a research field that holds substantial promise for improving trust and transparency and is recognized as a prerequisite for AI to continue making steady progress without disruption.

Through the lens of the existing literature, the survey provides an entry point for interested researchers and practitioners to learn key aspects of this young and rapidly growing body of research, reviewing existing approaches to the topic, discussing the trends surrounding its sphere, and presenting its major research trajectories. It became a widely cited reference organizing the emerging field of XAI.