III: Medium: Collaborative Research: Fairness in Web Database Applications

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

The Web has affected every corner of human life and society by providing the cyber-infrastructure to remove physical barriers between people. Myriad web database applications, including online recommendation systems, online shopping sites, location-based websites, resource-sharing platforms, social media, and service websites, have made people's lives unimaginably more connected, convenient, and cost-effective. The internal models, algorithms, ranking strategies, user profiling, and data resources shape the behavior of these web database applications. Unfortunately, these can include unfair practices that propagate, or even amplify, historical biases through their services, products, and recommendations. This project aims to detect such unfairness and to correct it where possible. A central question in detecting and correcting unfairness is how much knowledge can be assumed about these web systems' underlying data and algorithms. It is unrealistic to assume full knowledge, and it is hard even to detect unfairness without such assumptions. This project relies on making limited assumptions, such as the existence of a back-end database. Based on these, the project will detect unwarranted bias, develop responsible design tools one can use to avoid inadvertent unfairness, and implement third-party tools to tailor responses to reduce disparities between different demographic groups. 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.