Abstract:Deep learning models have revolutionized numerous fields, yet their decision-making processes often remain opaque, earning them the characterization of "black-box" models due to their lack of transparency and comprehensibility. This opacity presents significant challenges to understanding the rationale behind their decisions, thereby impeding their interpretability, explainability, and reliability. This review examines 718 studies published between 2015 and 2024 in high-impact journals indexed in SCI, SCI-E, SSCI, and ESCI, providing a crucial reference for researchers investigating methodologies and techniques in related domains. In this exploration, we evaluate a wide array of interpretability and explainability (XAI) strategies, including visual and feature-based explanations, local approach-based techniques, and Bayesian methods. These strategies are assessed for their effectiveness and applicability using a comprehensive set of evaluation metrics. Moving beyond traditional analyses, we propose a novel taxonomy of XAI methods, addressing gaps in the literature and offering a structured classification that elucidates the roles and interactions of these methods. Moreover, we explore the intricate relationship between interpretability and explainability, examining potential conflicts and highlighting the necessity for interpretability in practical applications. Through detailed comparative analysis, we underscore the strengths and limitations of various XAI methods across different data types, ensuring a thorough understanding of their practical performance and real-world utility. The review also examines model robustness against adversarial attacks, emphasizing the critical importance of transparency, reliability, and ethical considerations in model development. A significant emphasis is placed on identifying and mitigating biases in deep learning systems, providing insights into future research directions that aim to enhance fairness and reduce bias. By thoroughly reviewing current challenges and emerging research directions, this article equips researchers with the knowledge and tools to advance the development of more transparent, fair, and reliable deep learning systems. Ultimately, this work aims to bridge existing literature gaps by offering a forward-looking perspective that fosters innovation and progress in the field. This comprehensive review not only illuminates the current state of XAI methodologies but also contributes to the broader understanding and enhancement of deep learning systems, ensuring their ethical and equitable application across various domains.

Back to basics to open the black box

Black Box Optimization, Machine Learning, and No-Free Lunch Theorems

Open the Black Box Data-Driven Explanation of Black Box Decision Systems

Opening the Black Box of Neural Networks: Methods for Interpreting Neural Network Models in Clinical Applications

Cracking black-box models: Revealing hidden machine learning techniques behind their predictions

Open-environment Machine Learning.

Why black box machine learning should be avoided for high-stakes decisions, in brief

"How do I fool you?": Manipulating User Trust via Misleading Black Box Explanations

Beyond Average Performance -- exploring regions of deviating performance for black box classification models

Explainable Machine Learning: The Importance of a System-Centric Perspective

Clinical applications of machine learning algorithms: beyond the black box

Open-world Machine Learning: A Review and New Outlooks

Shedding Light on Black Box Machine Learning Algorithms: Development of an Axiomatic Framework to Assess the Quality of Methods that Explain Individual Predictions

Open Problems in Applied Deep Learning

Critical Empirical Study on Black-box Explanations in AI

A Grey-Box Ensemble Model Exploiting Black-Box Accuracy and White-Box Intrinsic Interpretability

Unlocking the black box: an in-depth review on interpretability, explainability, and reliability in deep learning

Foundations of Machine Learning

Classical Machine Learning: Seventy Years of Algorithmic Learning Evolution

Toward a `Standard Model' of Machine Learning

Benchmarking and survey of explanation methods for black box models