Fair AI for multi-agent decision making

Course Description
This seminar explores the principles and practices of ensuring fairness in algorithmic decision-making across various contexts, including resource allocation, classification, and recommendations. Through tutorials and paper presentations, the sessions will delve into different fairness concepts, techniques for designing algorithms with provable fairness guarantees, and tools for auditing and evaluating algorithms for fairness. This seminar is ideal for anyone interested in learning the tools and methods to design and assess fair solutions in complex multi-agent environments and excited about conducting research in this research space.  

Prerequisites

1. A course on Design and Analysis of Algorithms (16:198:513 or 01:198:344)

2. Introduction to Artificial Intelligence (01:198:440) or Machine Learning Principles (01:198:461) or a similar course.

Course Agenda

Week 1: Introduction to Fairness in AI

• Dive into the significance of fairness in AI systems.

• Explore key fairness concepts across multiple domains.

Week 2-3: Fairness in Resource Allocation

• Approaches from the social choice theory and key theorems

Week 4-5: Tackling Computational Challenges

• Uncover the challenges of ensuring fairness in machine learning –- recommenders and classifiers

• Review breakthrough results

Week 6: Auditing Fairness in Practice
Get hands-on with tools and methods for fairness auditing

Week 7-10: Student-led presentations

Present results and insights from the foundational papers in algorithmic fairness

Week 11-12: Final Project

• Group Presentations/Demo

• Report