Supervised Algorithmic Fairness in Distribution Shifts

Abstract

Supervised fairness-aware machine learning under distribution shifts is an emerging field that addresses the challenge of maintaining equitable and unbiased predictions when faced with changes in data distributions from source to target domains. In real-world applications, machine learning models are often trained on a specific dataset but deployed in environments where the data distribution may shift over time due to various factors. This shift can result in classifiers exhibiting poor generalization with low accuracy and making unfair predictions, which disproportionately impact certain groups identified by sensitive attributes, such as race and gender. 

In this tutorial, we begin by providing a comprehensive summary of various types of distribution shifts organized into two main categories and briefly discuss the factors contributing to the emergence of unfair outcomes by supervised learning models under such shifts. Then, we conduct a thorough examination of existing methods for maintaining algorithmic fairness based on these shifts and highlight six commonly used approaches in the literature. Furthermore, we introduce frameworks for fairness-aware generalization under distribution shifts, including the latest research developments. Finally, we explore the intersections with related research fields, address significant challenges, and propose potential directions for future studies.

Materials

• Tutorial Slides:

  • TBA

• Survey Paper:

  • Minglai Shao, Dong Li, Chen Zhao, Xintao Wu, Yujie Lin, Qin Tian. Supervised Algorithmic Fairness in Distribution Shifts: A Survey. In Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence (IJCAI), 2024.

  • Paper Link: https://www.ijcai.org/proceedings/2024/0909.pdf

Schedule

• Time: 4:15 pm - 6:00 pm (U.S. Eastern Time)

• Location: TBA

Outline

1. Part I: Introduction to Supervised Algorithmic Fairness

   • Real-world cases and issues of algorithmic fairness

   • Definition and importance of algorithmic fairness

   • Overview of different types of fairness in machine learning

   • Introduction to distribution shifts and their impact on fairness

2. Part II: Learning Fairness under Distribution Shifts

   • General settings and formulation of learning fairness under distribution shifts

   • Introduction of various types of distribution shifts 

   • Challenges posed by such shifts in maintaining fairness

3. Part III: Existing Approaches for Maintaining Algorithmic Fairness under Distribution Shifts

   • Introduction to methods for detecting and monitoring distribution shifts

   • Six commonly used strategies for adapting fairness to changing distributions

   • Introduction to past and latest related works within these six strategies

   • Introduction to benchmark datasets and evaluation metrics

4. Part IV: Future Directions, Related Research Fields, Conclusion

   • Challenges and potential future directions

   • Related research problems

   • Summary of key takeaways and insights from the tutorial

5. Part V: Q&A Session

   • Open floor for participants to ask questions and seek clarification

   • Discussion of specific challenges or scenarios raised by participants

   • Closing remarks and gratitude to participants for their engagement

Tutors' Bios

Past Tutorials:

• Chen Zhao, Xintao Wu. Supervised Algorithmic Fairness in Distribution Shifts. IEEE International Conference on Big Data (BigData'24), Washington D.C., USA, December, 15th, 2024. [Website]