Assessing and Mitigating Unfairness in AI Systems

Übersicht

Learn to assess and mitigate fairness issues in AI systems, focusing on healthcare disparities. Hands-on practice with Fairlearn library to evaluate and improve ML model performance across racial groups.

Lehrplan

• Introduction to Fairness in AI
Definition and importance of fairness in AI
Overview of fairness issues in AI systems
Case studies of bias and unfairness in AI
• Understanding Bias in Healthcare AI
Introduction to healthcare disparities
Common sources of bias in healthcare AI systems
The impact of unfair AI on racial groups in healthcare
• Fairness Metrics and Evaluation
Overview of fairness metrics
Selecting the right fairness metrics
Hands-on tutorial with Fairlearn: Calculating fairness metrics
• Introduction to the Fairlearn Library
Introduction and installation
Core functionalities of Fairlearn
Using Fairlearn in Python for model assessment
• Assessing Fairness in AI Models
Practical session: Evaluating a sample healthcare model
Using Fairlearn's dashboard for visualization
Interpreting fairness metrics results
• Techniques for Mitigating Unfairness
Pre-processing techniques
In-processing techniques
Post-processing techniques
Hands-on practice: Implementing mitigation strategies with Fairlearn
• Case Study: Improving Fairness in Healthcare Models
Analyzing a real-world healthcare model
Identifying bias and unfairness
Applying Fairlearn for bias mitigation
• Best Practices and Deployment
Strategies for maintaining fairness in deployed models
Continuous monitoring and feedback loops
Ethical considerations and regulatory compliance
• Capstone Project
Define a project using real or simulated healthcare data
Assess bias in the AI system
Apply mitigation strategies to improve model fairness
Present findings and solutions
• Conclusion and Future Directions
Recap of key learnings
Emerging trends in AI fairness
Resources for further learning and research

---

Fachgebiete

Conference Talks