III: Medium: Collaborative Research: Evaluating and Maximizing Fairness in Information Flow on Networks

University Of Colorado At Boulder, Boulder CO

Investigators

Abstract

Social networks (whom you know and whom you can reach) help determine access to hiring opportunities, education, and health information. They encode social capital based on network position, and in an era where access to information is crucial for advancement, this social capital can be immensely valuable. In this project, we study interventions on social networks that mitigate access inequality, that is, differences in access to information that emerge from where you are in the network. This project will develop novel mathematical and computational models to characterize how differences in position in a social network can amplify inequalities of access, and techniques to change the structure of the network that both increase the flow of information and reduce the overall inequities. Finally, the project will develop experimental methodology to assess the behavior of human agents in such online social networks to assess the validity of the designed interventions. The project will support the mentoring and training of underrepresented populations of undergraduate and graduate students, as well as the dissemination of the work through open-source software repositories and event organization at the top research venues in the field. We will develop mathematical and computational tools for the analysis of fairness in information access on networks. From there, we will characterize the algorithmic difficulty of mitigating information access gaps, develop efficient estimators to predict such gaps, and design intervention strategies to reduce these gaps. We will also consider how to characterize clusters of people who share similar access to information. More generally, we seek to connect the research on influence maximization to recent work on algorithmic fairness: the study of how automated procedures can perpetuate or exacerbate existing structural disadvantages of marginalized groups. The algorithms and results developed through these efforts will be evaluated using a combination of theoretical models, network repositories maintained by the PIs, real-world social network datasets, and experiments with volunteer participants. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.