CAREER: Public decision-making with crowdsourced data

Cornell University, Ithaca NY

Investigators

Abstract

Essential societal decisions and allocations rely on imperfect crowdsourced data. For example, in resident crowdsourcing, the public reports problems (for example, fallen trees and power lines and flooding after storms) that the government needs to address. State and national public health decisions (for example, allocation of vaccines and resources) rely on individual testing and reporting. However, decisions made without incorporating what is known regarding the deficiencies of crowdsourced data can be wasteful. For example, past work has shown that data from some areas may be systematically missing, leading to under-allocation of resources there. This research project will improve public decision-making by developing statistical methods to understand differential reporting behavior and engineer more efficient, transparent systems that account for missing information and heterogeneous behavior. This knowledge will help government and public-interest organizations allocate resources where they are most needed. This project will also educate data scientists and researchers for the public interest, including by providing continuing education and publicly available resources for municipal technology workers and open data hobbyists. This research pursues three objectives: 1) measuring statistical errors in report data, with a focus on public crowdsourcing of incidents and community health monitoring, where needs are spatially correlated and varied; 2) auditing responses to reports when resources are capacity constrained and multi-stage; 3) in collaboration with government and non-profit decision-makers, improving every stage of the response pipeline in practice. The research will contribute methods for general Bayesian inference, optimization, machine learning, and data-driven decision-making , applied to auditing and engineering systems in complex environments, including education, health, and government broadly. 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.