This book presents an overview and several applications of explainable artificial intelligence (XAI). It covers different aspects related to explainable artificial intelligence, such as the need to make the AI models interpretable, how black box machine/deep learning models can be understood using various XAI methods, different evaluation metrics for XAI, human-centered explainable AI, and applications of explainable AI in health care, security surveillance, transportation, among other areas.
The book is suitable for students and academics aiming to build up their background on explainable AI and can guide them in making machine/deep learning models more transparent. The book can be used as a reference book for teaching a graduate course on artificial intelligence, applied machine learning, or neural networks. Researchers working in the area of AI can use this book to discover the recent developments in XAI. Besides its use in academia, this book could be used by practitioners in AI industries, healthcare industries, medicine, autonomous vehicles, and security surveillance, who would like to develop AI techniques and applications with explanations.
Author(s): Mayuri Mehta, Vasile Palade, Indranath Chatterjee
Series: Intelligent Systems Reference Library, 232
Publisher: Springer
Year: 2022
Language: English
Pages: 272
City: Cham
Preface
Contents
Contributors
Abbreviations
1 Black Box Models for eXplainable Artificial Intelligence
1.1 Introduction to Machine Learning
1.1.1 Motivation
1.1.2 Scope of the Paper
1.2 Importance of Cyber Security in eXplainable Artificial Intelligence
1.2.1 Importance of Trustworthiness
1.3 Deep Learning (DL) Methods Contribute to XAI
1.4 Intrusion Detection System
1.4.1 Classification of Intrusion Detection System
1.5 Applications of Cyber Security and XAI
1.6 Comparison of XAI Using Black Box Methods
1.7 Conclusion
References
2 Fundamental Fallacies in Definitions of Explainable AI: Explainable to Whom and Why?
2.1 Introduction
2.1.1 A Short History of Explainable AI
2.1.2 Diversity of Motives for Creating Explainable AI
2.1.3 Internal Inconsistency of Motives for Creating XAI
2.1.4 The Contradiction Between the Motives for Creating Explainable AI
2.1.5 Paradigm Shift of Explainable Artificial Intelligence
2.2 Proposed AI Model
2.2.1 The Best Way to Optimize the Interaction Between Human and AI
2.2.2 Forecasts Are not Necessarily Useful Information
2.2.3 Criteria for Evaluating Explanations
2.2.4 Explainable to Whom and Why?
2.3 Proposed Architecture
2.3.1 Fitness Function for Explainable AI
2.3.2 Deep Neural Network is Great for Explainable AI
2.3.3 The More Multitasking the Better
2.3.4 How to Collect Multitasking Datasets
2.3.5 Proposed Neural Network Architecture
2.4 Conclusions
References
3 An Overview of Explainable AI Methods, Forms and Frameworks
3.1 Introduction
3.2 XAI Methods and Their Classifications
3.2.1 Based on the Scope of Explainability
3.2.2 Based on Implementation
3.2.3 Based on Applicability
3.2.4 Based on Explanation Level
3.3 Forms of Explanation
3.3.1 Analytical Explanation
3.3.2 Visual Explanation
3.3.3 Rule-Based Explanation
3.3.4 Textual Explanation
3.4 Frameworks for Model Interpretability and Explanation
3.4.1 Explain like I'm 5
3.4.2 Skater
3.4.3 Local Interpretable Model-Agnostic Explanations
3.4.4 Shapley Additive Explanations
3.4.5 Anchors
3.4.6 Deep Learning Important Features
3.5 Conclusion and Future Directions
References
4 Methods and Metrics for Explaining Artificial Intelligence Models: A Review
4.1 Introduction
4.1.1 Bringing Explainability to AI Decision—Need for Explainable AI
4.2 Taxonomy of Explaining AI Decisions
4.3 Methods of Explainable Artificial Intelligence
4.3.1 Techniques of Explainable AI
4.3.2 Stages of AI Explainability
4.3.3 Types of Post-model Explaination Methods
4.4 Metrics for Explainable Artificial Intelligence
4.4.1 Evaluation Metrics for Explaining AI Decisions
4.5 Use-Case: Explaining Deep Learning Models Using Grad-CAM
4.6 Challenges and Future Directions
4.7 Conclusion
References
5 Evaluation Measures and Applications for Explainable AI
5.1 Introduction
5.2 Literature Review
5.3 Basics Related to XAI
5.3.1 Understanding
5.3.2 Explicability
5.3.3 Explainability
5.3.4 Transparency
5.3.5 Explaining
5.3.6 Interpretability
5.3.7 Correctability
5.3.8 Interactivity
5.3.9 Comprehensibility
5.4 What is Explainable AI?
5.4.1 Fairness
5.4.2 Causality
5.4.3 Safety
5.4.4 Bias
5.4.5 Transparency
5.5 Need for Transparency and Trust in AI
5.6 The Black Box Deep Learning Models
5.7 Classification of XAI Methods
5.7.1 Global Methods Versus Local Methods
5.7.2 Surrogate Methods Versus Visualization Methods
5.7.3 Model Specific Versus Model Agnostic
5.7.4 Pre-Model Versus In-Model Versus Post-Model
5.8 XAI’s Evaluation Methods
5.8.1 Mental Model
5.8.2 Explanation Usefulness and Satisfaction
5.8.3 User Trust and Reliance
5.8.4 Human-AI Task Performance
5.8.5 Computational Measures
5.9 XAI’s Explanation Methods
5.9.1 Lime
5.9.2 Sp-Lime
5.9.3 DeepLIFT
5.9.4 Layer-Wise Relevance Propagation
5.9.5 Characteristic Value Evaluation
5.9.6 Reasoning from Examples
5.9.7 Latent Space Traversal
5.10 Explainable AI Stakeholders
5.10.1 Developers
5.10.2 Theorists
5.10.3 Ethicists
5.10.4 Users
5.11 Applications
5.11.1 XAI for Training and Tutoring
5.11.2 XAI for 6G
5.11.3 XAI for Network Intrusion Detection
5.11.4 XAI Planning as a Service
5.11.5 XAI for Prediction of Non-Communicable Diseases
5.11.6 XAI for Scanning Patients for COVID-19 Signs
5.12 Possible Research Ideology and Discussions
5.13 Conclusion
References
6 Explainable AI and Its Applications in Healthcare
6.1 Introduction
6.2 The Multidisciplinary Nature of Explainable AI in Healthcare
6.2.1 Technological Outlook
6.2.2 Legal Outlook
6.2.3 Medical Outlook
6.2.4 Ethical Outlook
6.2.5 Patient Outlook
6.3 Different XAI Techniques Used in Healthcare
6.3.1 Methods to Explain Deep Learning Models
6.3.2 Explainability by Using White-Box Models
6.3.3 Explainability Methods to Increase Fairness in Machine Learning Models
6.3.4 Explainability Methods to Analyze Sensitivity of a Model
6.4 Application of XAI in Healthcare
6.4.1 Medical Diagnostics
6.4.2 Medical Imaging
6.4.3 Surgery
6.4.4 Detection of COVID-19
6.5 Conclusion
References
7 Explainable AI Driven Applications for Patient Care and Treatment
7.1 General
7.2 Benefits of Technology and AI in Healthcare Sector
7.3 Most Common AI-Based Healthcare Applications
7.4 Issues/Concerns of Using AI in Health Care
7.5 Why Explainable AI?
7.6 History of XAI
7.7 Explainable AI’s Benefits in Healthcare
7.8 XAI Has Proposed Applications for Patient Treatment and Care
7.9 Future Prospects of XAI in Medical Care
7.10 Case Study on Explainable AI
7.11 Framework for Explainable AI
7.12 Conclusion
References
8 Explainable Machine Learning for Autonomous Vehicle Positioning Using SHAP
8.1 Introduction
8.1.1 Global Navigation Satellite System (GNSS) and Autonomous Vehicles
8.1.2 Navigation Using Inertial Measurement Sensors
8.1.3 Inertial Positioning Using Wheel Encoder Sensors
8.1.4 Motivation for Explainability in AV Positioning
8.2 eXplainable Artificial Intelligence (XAI): Background and Current Challenges
8.2.1 Why XAI in Autonomous Driving?
8.2.2 What is XAI?
8.2.3 Types of XAI
8.3 XAI in Autonomous Vehicle and Localisation
8.4 Methodology
8.4.1 Dataset: IO-VNBD (Inertial and Odometry Vehicle Navigation Benchmark Dataset)
8.4.2 Mathematical Formulation of the Learning Problem
8.4.3 WhONet’s Learning Scheme
8.4.4 Performance Evaluation Metrics
8.4.5 Training of the WhONet Models
8.4.6 WhONet’s Evaluation
8.4.7 SHapley Additive exPlanations (SHAP) Method
8.5 Results and Discussions
8.6 Conclusions
References
9 A Smart System for the Assessment of Genuineness or Trustworthiness of the Tip-Off Using Audio Signals: An Explainable AI Approach
9.1 Introduction
9.2 Background
9.3 Proposed Methodology
9.3.1 Dataset Used
9.3.2 Pre-processing
9.3.3 Feature Extracted
9.3.4 Feature Selected
9.3.5 Machine Learning in SER
9.3.6 Performance Index
9.4 Results and Discussion
9.5 Conclusion
References
10 Face Mask Detection Based Entry Control Using XAI and IoT
10.1 Introduction
10.2 Literature Review
10.3 Methodology
10.3.1 Web Application Execution
10.3.2 Implementation
10.3.3 Activation Functions
10.3.4 Raspberry Pi Webserver
10.4 Results
10.4.1 Dataset
10.4.2 Model Summary
10.4.3 Model Evaluation
10.5 Conclusion
References
11 Human-AI Interfaces are a Central Component of Trustworthy AI
11.1 Introduction
11.2 Regulatory Requirements for Trustworthy AI
11.3 Explicability—An Ethical Principle for Trustworthy AI
11.4 User-Centered Approach to Trustworthy AI
11.4.1 Stakeholder Analysis and Personas for AI
11.4.2 User-Testing for AI
11.5 An Example Use Case: Computational Pathology
11.5.1 AI in Computational Pathology
11.5.2 Stakeholder Analysis for Computational Pathology
11.5.3 Human-AI Interface in Computational Pathology
11.6 Conclusion
11.7 List of Abbreviations
References
