AI Bias Auditing and Explainability in Practice

• Chapter 1.1: What is AI Bias? Definitions and Categories
• Chapter 1.2: Sources of Bias in Datasets and Models
• Chapter 1.3: Social and Ethical Impacts of Algorithmic Bias
• Chapter 1.4: Case Studies in Healthcare, Finance, and HR

• Chapter 2.1: Why Explainability Matters in High-Stakes AI
• Chapter 2.2: Model Transparency vs. Post-Hoc Interpretability
• Chapter 2.3: Regulatory Expectations (EU AI Act, FTC, EEOC, etc.)
• Chapter 2.4: Trade-offs Between Accuracy and Interpretability

• Chapter 3.1: Fairness Metrics: Demographic Parity, Equal Opportunity, etc.
• Chapter 3.2: Tools for Bias Auditing: AIF360, Fairlearn, What-If Tool
• Chapter 3.3: Dataset Balancing and Preprocessing Techniques
• Chapter 3.4: Bias Mitigation During and After Training

• Chapter 4.1: Feature Importance and Global Model Insights
• Chapter 4.2: LIME, SHAP, and Anchors: Local Interpretability Methods
• Chapter 4.3: Explaining Black-Box Models vs. Interpretable ML
• Chapter 4.4: Generating and Presenting Explanations to Stakeholders

• Chapter 5.1: Building Ethical Guardrails for AI Systems
• Chapter 5.2: Model Cards and System Fact Sheets
• Chapter 5.3: Human-in-the-Loop Systems and Review Processes
• Chapter 5.4: Internal Accountability and Reporting Frameworks

• Chapter 6.1: Bias and Explainability in Real Products (Banking, Hiring, Healthcare)
• Chapter 6.2: Organizational Structures for Responsible AI
• Chapter 6.3: Capstone: Conduct a Bias + Explainability Audit of a Sample Model
• Chapter 6.4: Presenting Findings and Remediation Plans

Video content aligned with weekly modules

Foundations of AI Bias and Explainability

• Introduction to Bias in AI Systems
• Types of Bias: Data, Labeling, Sampling, and Outcome
• Ethical and Societal Risks of Algorithmic Discrimination
• Real-World Examples: Biased Models in Healthcare and Hiring
• Explainability vs. Interpretability: Definitions and Differences
• Why Explainability Matters: Legal, Ethical, and Technical Viewpoints
• Global Regulations Impacting AI Transparency (EU AI Act, EEOC, etc.)
• Transparency vs. Performance: Navigating the Trade-offs
• Week 1 Summary and Use Case Review

Techniques and Tools for Auditing and Explaining Models

• Understanding Fairness Metrics: Demographic Parity, Equalized Odds
• Bias Detection Tools: Intro to AIF360 and Fairlearn
• Bias Mitigation Techniques: Preprocessing, In-processing, Post-processing
• Data Balancing, Reweighting, and Sampling Strategies
• Feature Importance Techniques: Global vs. Local Explanations
• LIME and SHAP: How They Work and When to Use Them
• Hands-on with the What-If Tool and TensorBoard Fairness Indicators
• Comparing Interpretable Models vs. Black-Box Explainers
• Week 2 Case Lab: Auditing a Classifier on Real Data

Governance, Documentation, and Organizational Practice

• Creating Model Cards and Fact Sheets
• Designing Human-in-the-Loop Review Systems
• Internal AI Ethics Boards and Escalation Paths
• Reporting and Escalation in Regulated Industries
• Explainability for Non-Technical Stakeholders
• Real-World Case Study: Bias in a Loan Approval System
• Real-World Case Study: SHAP Interpretations for a Clinical Model
• Capstone Walkthrough: Auditing + Explaining a Sample Model
• Final Presentation Tips and Future Trends in Auditing and XAI

Title: Understanding and Framing AI Bias in Real-World Systems
Duration: 60 minutes
Focus: Core concepts of algorithmic bias, ethical impacts, and systemic risks
Guest: AI Ethics Researcher / Social Impact Technologist
Interactive: Group analysis of real bias incidents in lending, hiring, or healthcare

Title: Tools, Metrics, and Models: Inside a Bias Audit
Duration: 75 minutes
Focus: Walkthrough of fairness metrics, auditing workflows, and model explainers
Guest: Machine Learning Engineer / Responsible AI Tool Developer
Interactive: Live demo of SHAP and Fairlearn + participant Q&A on applying tools

Title: From Metrics to Accountability: Operationalizing Explainability
Duration: 90 minutes
Focus: Governance structures, human-in-the-loop systems, and documentation standards
Guest Panel: Data Scientist + AI Governance Lead + Compliance Manager
Interactive: Capstone feedback session + group discussion on presenting audit findings

Edit

Edit