LCN2 Seminar March 2023
- Friday 31 March 2023
- Room 312
59th LCN2 seminar
Speaker: Akrati Saxena (LIACS)
Title: Fairness-aware Link Prediction in Social Networks
In real-world complex networks, understanding the dynamics of their evolution has been of great interest to the scientific community. Link prediction, i.e., predicting non-existent but probable links, is an essential task in social network analysis (SNA), as the addition or removal of the links over time leads to network evolution. At the same time, in recent years, designing fairness-aware methods has received much attention in various domains, including machine learning, natural language processing, and information retrieval. A few famous examples are designing fair hiring systems, removing gender bias from word embedding, and creating a fair ranking for web search and advertisement systems. However, in social network analysis, understanding the impact of structural biases and inequalities of social systems on the fairness and accuracy of different methods (such as link prediction, influence maximization, and community detection methods) has not received much attention. In this talk, I will first discuss link prediction methods and highlight how the structural bias of social networks affects the fairness of link prediction. It will be followed by an explanation of fairness constraints that can be considered while designing link prediction methods. Next, I will discuss heuristic and network embedding based methods that mitigate the impact of these biases to predict fair and diverse links in a network. The talk will be concluded with open research directions and our ideas to bridge the gap between fairness and SNA.
Afterwards there will be drinks in the Foobar.