Explainable artificial intelligence: a comprehensive review

Abstract

Thanks to the exponential growth in computing power and vast amounts of data, artificial intelligence (AI) has witnessed remarkable developments in recent years, enabling it to be ubiquitously adopted in our daily lives. Even though AI-powered systems have brought competitive advantages, the black-box nature makes them lack transparency and prevents them from explaining their decisions. This issue has motivated the introduction of explainable artificial intelligence (XAI), which promotes AI algorithms that can show their internal process and explain how they made decisions. The number of XAI research has increased significantly in recent years, but there lacks a unified and comprehensive review of the latest XAI progress. This review aims to bridge the gap by discovering the critical perspectives of the rapidly growing body of research associated with XAI. After offering the readers a solid XAI background, we analyze and review various XAI methods, which are grouped into (i) pre-modeling explainability, (ii) interpretable model, and (iii) post-modeling explainability. We also pay attention to the current methods that dedicate to interpret and analyze deep learning methods. In addition, we systematically discuss various XAI challenges, such as the trade-off between the performance and the explainability, evaluation methods, security, and policy. Finally, we show the standard approaches that are leveraged to deal with the mentioned challenges.

Explainable Automated Recognition of Emotional States from Canine Facial Expressions: The Case of Positive Anticipation and Frustration

“…62 . These topics are fundamental in AI, and are addressed by a huge body of research [63][64][65] , with the majority of efforts focusing on deep learning approaches, whose interpretability is limited by their complex structure 66 . Explainability methods are by their nature domain-specific: providing explanations for automated personality trait recognition in job interviews is different, e.g., from providing clinical justification for medical decisions 62 .…”

Emerging early diagnostic methods for acute kidney injury

“…Most of these models are black-box predictions that cannot readily be explained to clinicians. Transparency and interpretability are necessary for the widespread introduction of artificial intelligence models into clinical practice 63 - 65 . The above-mentioned weak classifier ranking and shape interpretation methods not consider the dependence between variables, which inevitably lead to the correlation bias.…”

Explanatory machine learning for sequential human teaching

“…Transparency typically refers to the ability of a model or a system to be human-understandable on its own [11]. Interpretability relates to the degree of clarity of information revealed to the user by a system [62]. Explainability stresses the role of a system as an "explainer" interface to the user [11].…”

“…An intelligent machine can in this way fulfil an integrated role that contributes to both the "stock of software" and the "stock of knowledge". Owing to increasing awareness of the importance of explainable AI, reported by comprehensive reviews [9,1,62], a great number of studies have emerged with an emphasis on systems that provide support for human understanding. In addition, there is an increasing emphasis on AI that interacts to keep humans "in the loop" [61,79].…”