Fairness is becoming a paramount consideration for data scientists. Mounting evidence indicates that the widespread deployment of machine learning and AI in business and government is reproducing the same biases we're trying to fight in the real world. But what does fairness mean when it comes to code? This practical book covers basic concerns related to data security and privacy to help data and AI professionals use code that's fair and free of bias.
Many realistic best practices are emerging at all steps along the data pipeline today, from data selection and preprocessing to closed model audits. Author Aileen Nielsen guides you through technical, legal, and ethical aspects of making code fair and secure, while highlighting up-to-date academic research and ongoing legal developments related to fairness and algorithms.
• Identify potential bias and discrimination in data science models
• Use preventive measures to minimize bias when developing data modeling pipelines
• Understand what data pipeline components implicate security and privacy concerns
• Write data processing and modeling code that implements best practices for fairness
• Recognize the complex interrelationships between fairness, privacy, and data security created by the use of machine learning models
• Apply normative and legal concepts relevant to evaluating the fairness of machine learning models
Author(s): Aileen Nielsen
Edition: 1
Publisher: O'Reilly Media
Year: 2020
Language: English
Commentary: Vector PDF
Pages: 346
City: Sebastopol, CA
Tags: Equity; Machine Learning; Security; Ethics; Python; Law; Privacy; Deployment; Best Practices; Auditing; Explainability
Copyright
Table of Contents
Preface
Goals of This Book
Practical Notes on the Book
Conventions Used in This Book
Using Code Examples
O’Reilly Online Learning
How to Contact Us
Acknowledgments
Chapter 1. Fairness, Technology, and the Real World
Fairness in Engineering Is an Old Problem
Our Fairness Problems Now
Community Norms
Equity and Equality
Security
Privacy
Legal Responses to Fairness in Technology
The Assumptions and Approaches in This Book
What If I’m Skeptical of All This Fairness Talk?
Won’t Fairness Slow Down Innovation?
Are There Any Real-World Consequences for Not Developing Fairness-Aware Practices?
What Is Fairness?
Rules to Code By
Equality and Equity
Security
Privacy
Chapter 2. Understanding Fairness and the Data Science Pipeline
Metrics for Fairness
Measures of Equity
Measures of Privacy
Measures of Security
Connected Concepts
Privacy and Security
Privacy and Equity
Equality and Security
Accuracy and Fairness
Automated Fairness?
Checklist of Points of Entry for Fairness in the Data Science Pipeline
Assembling a Data Set
Modeling
Interface
Concluding Remarks
Chapter 3. Fair Data
Ensuring Data Integrity
True Measurements
Proportionality and Sampling Technique
Choosing Appropriate Data
Equity
Privacy
Security
Case Study: Choosing the Right Question for a Data Set and the Right Data Set for a Question
Quality Assurance for a Data Set: Identifying Potential Discrimination
A Timeline for Fairness Interventions
Comprehensive Data-Acquisition Checklist
Concluding Remarks
Chapter 4. Fairness Pre-Processing
Simple Pre-Processing Methods
Suppression: The Baseline
Massaging the Data Set: Relabeling
AIF360 Pipeline
Loading the Data
Fairness Metrics
The US Census Data Set
Suppression
Reweighting
How It Works
Code Demonstration
Learning Fair Representations
How It Works
Code Demonstration
Optimized Data Transformations
How It Works
Code Demonstration
Fairness Pre-Processing Checklist
Concluding Remarks
Chapter 5. Fairness In-Processing
The Basic Idea
The Medical Data Set
Prejudice Remover
How It Works
Code Demonstration
Adversarial Debiasing
How It Works
Code Demonstration
In-Processing Beyond Antidiscrimination
Model Selection
Concluding Remarks
Chapter 6. Fairness Post-Processing
Post-Processing Versus Black-Box Auditing
The Data Set
Equality of Opportunity
How It Works
Code Demonstration
Calibration-Preserving Equalized Odds
How It Works
Code Demonstration
Concluding Remarks
Chapter 7. Model Auditing for Fairness and Discrimination
The Parameters of an Audit
Scoping: What Should We Audit?
Black-Box Auditing
Running a Model Through Different Counterfactuals
Model of the Model
Auditing Black-Box Models for Indirect Influence
Concluding Remarks
Chapter 8. Interpretable Models and Explainability Algorithms
Interpretation Versus Explanation
Interpretable Models
GLRM: How It Works
Code Demonstration
Explainability Methods
SHAP and LIME: The Workhorses for Local Post Hoc Explanations
Data-Driven Explanation
Explainability Metrics
What Interpretation and Explainability Miss
Attacks on Explainable Machine Learning
Interpretation and Explanation Checklist
Concluding Remarks
Chapter 9. ML Models and Privacy
Membership Attacks
How It Works
Code Demonstration
Other Privacy Problems and Attacks
Important Privacy Techniques
Concluding Remarks
Chapter 10. ML Models and Security
Evasion Attacks
How It Works
Code Demonstration
Defending Against Adversarial Attacks
Some Evasion Attack Packages
Why Do Evasion Attacks Matter to You?
Poisoning Attacks
How They Work
Defenses Against Poisoning Attacks
Some Poisoning Attack Packages
Why Do Poisoning Attacks Matter to You?
Concluding Remarks
Chapter 11. Fair Product Design and Deployment
Reasonable Expectations
Expectations of Moving Targets
Clear Communication
Fiduciary Obligations
Respecting Traditional Spheres of Privacy and Private Life
Value Creation
Complex Systems
The Impact of the Product Life Cycle
The Need for Record Keeping
The Need for Experts
Clear Security Promises and Delineated Limitations
Reasonable Expectations of Security
Possibility of Downstream Control and Verification
Verification Systems and Obligations
Product Iteration Timelines
Tracking Downstream Users
Products That Work Better for Privileged People
Dark Patterns
Fair Products Checklist
Concluding Remarks
Chapter 12. Laws for Machine Learning
Personal Data
GDPR
California Consumer Privacy Act
Data Broker Laws
Algorithmic Decision Making
GDPR
Proposed US Laws for Algorithms
Security
HIPAA
FTC Guidance on Cybersecurity
Tort Law
Logical Processes
Right to an Explanation
Freedom of Information Laws
Due Process
Some Application-Specific Laws
Biometrics
Local Ordinances on Facial Recognition
Chat Bots
Concluding Remarks
Index
About the Author
Colophon
