HARNESSING THE EAND#64256ECTS OF ALGORITHMIC BIAS IN ONLINE SOCIAL NETWORKS
Abstract
Modern societies share ideas, engage in cross-cultural cooperation, and participate in democratic processes within a pervasive, highly variable, and mostly hidden ecosystem of content filtering technologies that determine what information we see and share online. Despite the extraordinary impact of these algorithms on our daily lives and on the functioning of society as a whole, little is publicly known about their proprietary components, or the effect of the associated bias on web-driven processes of opinion formation, social influence and information transfer. It is thus unclear whether algorithmic bias has a harmful influence on collective opinion formation and decision-making, or whether it has a role (and if so, to what extent) in the increasingly polarized political debate of our day. The goal of the present proposal is to understand the impact of algorithmic bias on society and to develop a flexible modeling framework that combines plausible filtering algorithms with mechanisms of human interaction over dynamic social networks. Our driving hypothesis is that causal links between algorithms and social behavior can be discovered via model-based data analysis of popular online media platforms. Our aim is to assess the large-scale effect of algorithmic bias on society by modeling its interplay with the individual decisions of many, and thus arrive at heuristics to potentially correct the most harmful consequences of online bias. Our work has the potential of paving the road for much needed, scientifically grounded regulation of online ecosystems, informing the decision-making process of platform designers, policy makers, and government agencies.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 11, 2021
- Source ID
- FA86552017020
Entities
People
- Gerardo Iñiguez
Organizations
- Air Force Office of Scientific Research
- Central European University
- United States Air Force