A Review of Explainable Artificial Intelligence

Artificial intelligence (AI) has caused revolutionary changes in many areas of our lives. Especially deep learning-based models introduced in recent years have turned into complex black box algorithms whose decisionmaking processes and working mechanisms are difficult to understand, focusing on prediction performance. The black box problem causes some problems in terms of social acceptance and fair and ethical use of artificial intelligence systems. To address these problems and increase trust in artificial intelligence models, the idea of Explainable Artificial Intelligence (XAI) was introduced. XAI aims to create an infrastructure where even non-technical users can easily understand the functioning and decisions of AI models by increasing the transparency, reliability and fairness of AI. XAI has become a popular research topic in the field of artificial intelligence in recent years. In this study, the developing field of XAI has been examined in general, evaluated from different perspectives, and a detailed definition, technique and terminology explanation has been brought together. We hope that this study will guide researchers in the development of safer, fairer and more transparent XAI techniques by providing an evaluation of the field of XAI from different perspectives.