Like other powerful technologies, AI and machine learning present significant opportunities. To reap the full benefits of ML, organizations must also mitigate the considerable risks it presents. This report outlines a set of actionable best practices for people, processes, and technology that can enable organizations to innovate with ML in a responsible manner.
Authors Patrick Hall, Navdeep Gill, and Ben Cox focus on the technical issues of ML as well as human-centered issues such as security, fairness, and privacy. The goal is to promote human safety in ML practices so that in the near future, there will be no need to differentiate between the general practice and the responsible practice of ML.
This report explores:
- People: Humans in the Loop--Why an organization's ML culture is an important aspect of responsible ML practice
- Processes: Taming the Wild West of Machine Learning Workflows--Suggestions for changing or updating your processes to govern ML assets
- Technology: Engineering ML for Human Trust and Understanding--Tools that can help organizations build human trust and understanding into their ML systems
- Actionable Responsible ML Guidance--Core considerations for companies that want to drive value from ML
- Title
- Responsible Machine Learning
- Subtitle
- Actionable Strategies for Mitigating Risks and Driving Adoption
- Publisher
- O'Reilly Media
- Author(s)
- Benjamin Cox, Navdeep Gill, Patrick Hall
- Published
- 2020-10-02
- Edition
- 1
- Format
- eBook (pdf, epub, mobi)
- Pages
- 69
- Language
- English
- ISBN-13
- 9781492090847
- License
- Compliments of H2O.ai
Preface
1. Introduction to Responsible Machine Learning
What Is Responsible Machine Learning?
Responsible Artificial Intelligence
A Responsible Machine Learning Definition
2. People: Humans in the Loop
Responsible Machine Learning Culture
Accountability
Dogfooding
Demographic and Professional Diversity
Cultural Effective Challenge
Going Fast and Breaking Things
Get in the Loop
Human Audit of Machine Learning Systems
Domain Expertise
User Interactions with Machine Learning
User Appeal and Operator Override
Kill Switches
Going Nuclear: Public Protests, Data Journalism, and White-Hat Hacking
3. Processes: Taming the Wild West of Machine Learning Workflows
Discrimination In, Discrimination Out
Algorithmic Discrimination and US Regulations
Data Privacy and Security
Machine Learning Security
Legality and Compliance
Model Governance
Model Monitoring
Model Documentation
Hierarchy and Teams for Model Governance
Model Governance for Beginners
AI Incident Response
Organizational Machine Learning Principles
Corporate Social Responsibility and External Risks
Corporate Social Responsibility
Mitigating External Risks
4. Technology: Engineering Machine Learning for Human Trust and Understanding
Reproducibility
Metadata
Random Seeds
Version Control
Environments
Hardware
Interpretable Machine Learning Models and Explainable AI
Interpretable Models
Post hoc Explanation
Model Debugging and Testing Machine Learning Systems
Software Quality Assurance for Machine Learning
Specialized Debugging Techniques for Machine Learning
Benchmark Models
Model Debugging
Model Monitoring
Discrimination Testing and Remediation
Testing for Discrimination
Remediating Discovered Discrimination
Securing Machine Learning
Machine Learning Attacks
Countermeasures
Privacy-Enhancing Technologies for Machine Learning
Federated Learning
Differential Privacy
Causality
5. Driving Value with Responsible Machine Learning Innovation
Trust and Risk
Signal and Simplicity
The Future of Responsible Machine Learning
Further Reading
AcknowledgmentsRelated Books
