Causality in the Real World: From Geometry to Privacy Foundations
Abstract
Abstract Approved for Public Released Causal inference is critical in almost all areas of policy making, from medicine to politicalsciences and people management/deployment. This project ad- dresses two relatively understudied directions in the field. First, we propose to investigate how privacy interacts with causal discovery from observational data. While tools from differential privacy are ubiquitous, and virtually all properly protected data is affected by them, their impact in causal discovery is still in its infancy. Studying this, starting from standard tools such as re-gression discontinuity and synthetic controls, is the first part of the proposed project. The second part addresses potential connections between geometry and causal inference. In particular, how curvature in networks affects it. This opens a new direction of research where geometric and topological concepts are brought to discovery notonly causal variables but also the confidence in their estimation. Both components of the project come together when considering privacy tools in networks, and also when addressing the confidence and accuracy of causal discovery algorithms.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 11, 2024
- Source ID
- N000142412211
Entities
People
- Guillermo Sapiro
Organizations
- Duke University
- Office of Naval Research
- United States Navy