AI Ethics Resource Collection

Introduction

The rapid advancement of Artificial Intelligence (AI) has brought about numerous opportunities and challenges. As AI systems become increasingly pervasive in our lives, it is crucial to ensure that they are developed and deployed ethically and responsibly. This collection of resources on AI Ethics provides a comprehensive understanding of the ethical considerations and implications surrounding AI technologies. When looking through this collection, you will explore the ethical frameworks, principles, and guidelines necessary to navigate the complex landscape of AI development and usage.

Objectives

• Understand the fundamental concepts and principles of AI ethics.

• Explore the ethical challenges and dilemmas associated with AI technologies and applications.

• Understand industry standards and frameworks for implementing AI Ethics through out the Model Development Lifecycle

• Engage in case studies analysis to explore real-world ethical challenges in AI.

• Gain hands-on practical knowledge in achieving the AI Ethics principles during the model development.

Who is this for?

This collection of resources on AI Ethics is designed to cater to a broad audience with diverse interests and backgrounds in AI. It is ideal for:

1. AI Practitioners and Engineers: Professionals working directly with AI systems such as Machine Learning Engineers, Data Scientists, AI Researchers, and Software Engineers will find this collection extremely beneficial. It will provide them with a robust understanding of the ethical implications of their work and equip them with practical tools and methodologies to build responsible AI systems.

2. Leaders and Decision-makers: Executives, Managers, and Leaders who are involved in the strategic decision-making process concerning the implementation and use of AI within their organizations will gain significant insights. The collection offers a deep understanding of the ethical risks and responsibilities associated with AI, and can help to shape ethical AI strategy and governance.

3. AI Enthusiasts and the Public: Anyone interested in AI, its societal implications, and ethical considerations will find useful materials in this collection. The blend of technical and non-technical content makes it accessible for individuals regardless of their technical proficiency.

Resources

(please note that each resource contains a link external to the Diverse AI website that will open in a new window)

Figure 1:AI Guiding Principles - (Source: Khan 2018 - Devopedia.org)
Devopedia. 2022. "Ethical AI." Version 6, February 15. Accessed 2023-05-02. https://devopedia.org/ethical-ai

Introduction to Ethics and AI

• What is Automated Decision Making (A Chapter from the FAIRML Handbook)

• Understanding ethics and its relevance to AI (A big picture from IBM)

• Understanding ethics and its relevance to AI (Deep Dive Academic Guide from Stanford University)

• OECD’s Principles of Ethical AI - An Overview
  

Key Considerations for Ethics in AI

• Bias and Fairness in AI - An Overview (Article by Meghan Reagan PHD)

• Explainability & Transparency (Reading from the Royal Society)

• Privacy & Robustness (Reading From IBM)

Understanding Bias and Unfairness in AI

This is a comprehensive exploration into the complex realm of bias and fairness as they intersect with artificial intelligence. As AI systems increasingly permeate our daily lives, influencing everything from online search results to loan approvals, the issues of bias and fairness within these systems have become more critical than ever. This section aims to equip you with a deep understanding of these concepts, showcasing how they manifest within AI algorithms, the implications for individuals and society, and strategies to mitigate their effects. It's an essential collection for anyone interested in ensuring our AI-driven future is equitable, just, and truly beneficial for all.

Papers and articles

• Definitions of Bias (Academic Paper from Eirini et al)

• Definition of Fairness (A Chapter from the FAIRML Handbook)

• Sources of Harm In the Machine Learning Development Lifecycle (Academic Paper from MIT)

• Fairness Metrics for Mitigating biases and unfairness in AI systems (Blog Post from Hackernoon)

• Toolkits for evaluating Bias and Fairness in Practice (Reading by Edwin Wenink)

Case Studies

• Case Study 1: Credit Risk Scoring System (From the FairML Book)

• Case Study 2: Predictive Policing (A Legal Case from Fair Trials)

• Case Study 3: Predictive Racial Profiling with the COMPAS Dataset (A Case Study From Propublica)

Activities and exercises

• Bias Reduction Strategies

• Fairness Aware Data Mining (A Tutorial by Toshiro Kamashima)

• Achieving Fairness in AI Models (Technical Paper from Roh et al)

• Fairness in Racial Profiling in the criminal justice system (Hands on Lab from Stanford University)

• Debiasing Facial Recognition Systems (Hands on Lab from MIT)

Courses

• Ethics of Artificial Intelligence (A Course from Stanford University’s Human AI Institute)

Books

• The FairML Book (A Textbook from Borocas et al)

Understanding Explainability and Transparency in AI

The rapidly evolving landscape of Artificial Intelligence (AI) has also introduced both ground-breaking opportunities and new challenges, making the principles of explainability and transparency in AI systems more critical than ever. "Understanding Explainability and Transparency in AI Systems," aims to demystify these complex concepts, making them comprehensible and practical. As AI models become increasingly intricate and their decision-making processes less transparent, understanding why and how these models make certain decisions is fundamental for trust, ethics, and legal compliance. This section delves into the heart of AI systems, exploring the mechanisms that drive their decisions, and equip you with the knowledge and skills to interpret, explain, and communicate these decisions effectively, fostering trust and ensuring responsible AI deployment.

Papers and articles

• Explainable AI: The Basics (Essential Reading from Royal Society)

• Principles of Explainable AI (Article from the National Institute of Standards and Technology)

• Case Study 1

• Tools For Explainable AI (Article from Edwin Weninck)

• More Explainability tools (Article from AI4People)

• Explainable AI In Machine Learning (Technical Paper From Gohel et al)

• The Shapley Value (Technical Paper from Benedek et al)

• The Lime Value (Technical Paper from Rebiero et al)

Case Studies

• Case Studies in Explainability (Write up from Zevenbergen et al)

Activities and exercises

• An Introduction to Explainable AI With Shapley Values (Lab Tutorial From Slundberg)

• Using Lime for Explainable AI (Lab tutorial from Geeks For Geeks)

• How to Explain an Image Classifier using LIME (Lab Tutorial From Towards Data Science)

Courses

• Explainable AI Using Python (A course from Udemy)

Books

• Interpretable Machine Learning (A Textbook from Christoph Monar)

Understanding Privacy and Robustness in AI Systems

Understanding Privacy and Robustness in AI Systems is a comprehensive exploration into the critical dimensions of privacy and robustness in the rapidly evolving field of artificial intelligence. As AI systems increasingly integrate into our everyday lives, understanding how to maintain user privacy and ensure system robustness has never been more paramount. This section provides an in-depth look into the current challenges and best practices for achieving privacy and robustness in AI systems, providing you with the tools and insights needed to develop, manage, and evaluate AI applications responsibly and securely. Delving into key topics such as data protection, algorithmic transparency, adversarial attacks, and system resilience.

Papers and articles

• Understanding Privacy in AI Systems (Article from Ericsson)

• Understanding Robustness in AI Systems (Article from IBM)

• Understanding Data Drift for AI Robustness (Article from Data Camp)

• Privacy Preserving Learning Framework (Academic Paper from Riffel et al)

• Differential Privacy for Data Protection During Training (A Technical Overview from Harvard University)

• Tools for AI Robustness (Article from Edwin Weninck)

Case Studies

• Data Privacy Breach in US EdTech Industry (A Case Study from The Markup)

• Case Studies in AI Privacy (Article from IAPP)

• Case Study in ML Robustness (Academic Paper from IEEE)

Activities and exercises

• Calculating ML/AI Data Drift in Python (Article from Vatsal on Towards Data Science)

• Secure and Private AI (Hands on Tutorial from udacity)

• Introduction to Adversarial Attacks on Machine Learning models (Hands on Tutorials from NTU)

• Introduction to Federated Learning for Privacy Preserving Machine Learning (Article from Saheed Tijani)