FARA: Future-Aware Ranking Algorithm for Fairness Optimization

By Tao Yang et al

Published on Oct. 21, 2023

Read the original document by opening this link in a new tab.

Table of Contents

1. INTRODUCTION
2. RELATED WORK
3. BACKGROUND AND PRIOR KNOWLEDGE
4. PROPOSED METHOD
4.1 Future-aware Ranking Objective
4.2 Joint Optimization of Ranklists
4.3 Vertical Allocation Method
5. THEORETICAL ANALYSIS
6. EMPIRICAL RESULTS
7. CONCLUSION
8. REFERENCES

Summary

The paper introduces FARA, a Future-Aware Ranking Algorithm for Fairness Optimization, which aims to optimize ranking relevance and fairness in ranking systems. FARA plans ahead by jointly optimizing multiple ranklists for future sessions, unlike existing greedy algorithms. The proposed method involves a two-phase solution path: planning future exposure and constructing optimal ranklists. By using a Taylor series expansion, FARA estimates future fairness based on exposure increments to maximize fairness. The paper provides theoretical analysis and empirical results demonstrating the effectiveness of FARA compared to existing fair ranking algorithms.