Responsible AI: Best Practices for Creating Trustworthy AI Systems

Description

THE FIRST PRACTICAL GUIDE FOR OPERATIONALIZING RESPONSIBLE AIFROM MUL TI°LEVEL GOVERNANCE MECHANISMS TO CONCRETE DESIGN PATTERNS AND SOFTWARE ENGINEERING TECHNIQUES.

AI is solving real-world challenges and transforming industries. Yet, there are serious concerns about its ability to behave and make decisions in a responsible way. Operationalizing responsible AI is about providing concrete guidelines to a wide range of decisionmakers and technologists on how to govern, design, and build responsible AI systems. These include governance mechanisms at the industry, organizational, and team level; software engineering best practices; architecture styles and design patterns; system-level techniques connecting code with data and models; and trade-offs in design decisions. Responsible AI includes a set of practices that technologists (for example, technology-conversant decision-makers, software developers, and AI practitioners) can undertake to ensure the AI systems they develop or adopt are trustworthy throughout the entire lifecycle and can be trusted by those who use them. The book offers guidelines and best practices not just for the AI part of a system, but also for the much larger software infrastructure that typically wraps around the AI.

• First book of its kind to cover the topic of operationalizing responsible AI from the perspective of the entire software development life cycle.
• Concrete and actionable guidelines throughout the lifecycle of AI systems, including governance mechanisms, process best practices, design patterns, and system engineering techniques.
• Authors are leading experts in the areas of responsible technology, AI engineering, and software engineering.
• Reduce the risks of AI adoption, accelerate AI adoption in responsible ways, and translate ethical principles into products, consultancy, and policy impact to support the AI industry.
• Online repository of patterns, techniques, examples, and playbooks kept up-to-date by the authors.
• Real world case studies to demonstrate responsible AI in practice.
• Chart the course to responsible AI excellence, from governance to design, with actionable insights and engineering prowess found in this defi nitive guide.

Sample Content

Table of Contents

    Preface.. . . . . . . . . . . . . . . . . xv

    About the Author.. . . . . . . . . . . . . . xix

Part I Background and Introduction. . . . . . . . . . . . .1

1 Introduction to Responsible AI. . . . . . . . . 3

    What Is Responsible AI?. . . . . . . . . . . . 4

    What Is AI?. . . . . . . . . . . . . . 6

    Developing AI Responsibly: Who Is Responsible for Putting the

    Responsible into AI?.. . . . . . . . . . . . 8

    About This Book.. . . . . . . . . . . . . 9

    How to Read This Book.. . . . . . . . . . . . 11

2 Operationalizing Responsible AI: A Thought ExperimentRobbie the Robot.. . . . . . . . 13

    A Thought ExperimentRobbie the Robot.. . . . . . . . 13

    Summary. . . . . . . . . . . . . . 22

Part II Responsible AI Pattern Catalogue. . . . . . . . . . . 23

3 Overview of the Responsible AI Pattern Catalogue. . . . . 25

    The Key Concepts.. . . . . . . . . . . . . 25

    Why Is Responsible AI Different?. . . . . . . . . . 30

    A Pattern-Oriented Approach for Responsible AI.. . . . . . . 32

4 Multi-Level Governance Patterns for Responsible AI.. . . . 39

    Industry-Level Governance Patterns. . . . . . . . . 42

    Organization-Level Governance Patterns.. . . . . . . . 56

    Team-Level Governance Patterns.. . . . . . . . . . 72

    Summary. . . . . . . . . . . . . . 85

5 Process Patterns for Trustworthy Development Processes. . . 87

    Requirements.. . . . . . . . . . . . . 88

    Design. . . . . . . . . . . . . . . 96

    Implementation.. . . . . . . . . . . . . 105

    Testing. . . . . . . . . . . . . . . 110

    Operations. . . . . . . . . . . . . . 114

    Summary. . . . . . . . . . . . . . 120

6 Product Patterns for Responsible-AI-by-Design.. . . . . 121

    Product Pattern Collection Overview.. . . . . . . . . 122

    Supply Chain Patterns. . . . . . . . . . . . 123

    System Patterns. . . . . . . . . . . . . 134

    Operation Infrastructure Patterns. . . . . . . . . 141

    Summary. . . . . . . . . . . . . . 158

7 Pattern-Oriented Reference Architecture for Responsible-AI-by-Design. . . . . . . . . 159

    Architectural Principles for Designing AI Systems. . . . . . 160

    Pattern-Oriented Reference Architecture.. . . . . . . . 161

    Summary. . . . . . . . . . . . . . 165

8 Principle-Specific Techniques for Responsible AI.. . . . . 167

    Fairness.. . . . . . . . . . . . . . 167

    Privacy. . . . . . . . . . . . . . . 172

    Explainability. . . . . . . . . . . . . 178

    Summary. . . . . . . . . . . . . . 182

Part III Case Studies. . . . . . . . . . . . . . .  183

9 Risk-Based AI Governance in Telstra. . . . . . . 185

    Policy and Awareness.. . . . . . . . . . . . 186

    Assessing Risk.. . . . . . . . . . . . . 188

    Learnings from Practice. . . . . . . . . . . 192

    Future Work. . . . . . . . . . . . . . 195

10 Reejig: The Worlds First Independently Audited Ethical Talent AI.. . . . . . . . . . . 197

    How Is AI Being Used in Talent?.. . . . . . . . . . 198

    What Does Bias in Talent AI Look Like?.. . . . . . . . 200

    Regulating Talent AI Is a Global Issue.. . . . . . . . . 201

    Reejigs Approach to Ethical Talent AI. . . . . . . . . 202

    How Ethical AI Evaluation Is Done: A Case Study in Reejigs World-First Independently Audited Ethical Talent AI. . . . . . . . 204

    Overview.. . . . . . . . . . . . . 204

    Project Overview. . . . . . . . . . . . . 206

    The Ethical AI Framework Used for the Audit.. . . . . . . 207

    The Benefits of Ethical Talent AI.. . . . . . . . . . 210

    Reejigs Outlook on the Future of Ethical Talent AI.. . . . . . 211

11 Diversity and Inclusion in Artificial Intelligence.. . . . . 213

    Importance of Diversity and Inclusion in AI.. . . . . . . 215

    Definition of Diversity and Inclusion in Artificial Intelligence. . . . 216

    Guidelines for Diversity and Inclusion in Artificial Intelligence. . . . 219

    Conclusion.. . . . . . . . . . . . . . 234

Part IV Looking to the Future. . . . . . . . . . . . . 237

12 The Future of Responsible AI.. . . . . . . . . 239

    Regulation. . . . . . . . . . . . . . 241

    Education.. . . . . . . . . . . . . . 242

    Standards.. . . . . . . . . . . . . . 244

    Tools.. . . . . . . . . . . . . . . 245

    Public Awareness.. . . . . . . . . . . . 246

    Final Remarks.. . . . . . . . . . . . . 246

Part V Appendix. . . . . . . . . . . . . . . . 249

9780138073923, TOC, 11/7/2023

Updates

Submit Errata