On Dyadic Fairness: Exploring and Mitigating Bias in Graph Connections

Anonymous

Keywords: graphs, fairness, representation learning

Abstract: This blog post discusses the ICLR 2021 paper "On Dyadic Fairness: Exploring and Mitigating Bias in Graph Connections" by Li et al., highlighting the importance of its theoretical results while critiquing the notions and applications of dyadic fairness provided. This paper presents a beautiful proof that can be followed to analyze representation disparities produced by various message-passing algorithms, and an algorithmic skeleton for improving the fairness of many message-passing algorithms. At the same time, it is essential that, as a community, we critically analyze for which applications a fair link prediction algorithm can distribute justice and contextualize our understandings of the politics and limitations of different notions of fairness in link prediction applications.

ICLR Paper: https://openreview.net/forum?id=xgGS6PmzNq6