Monitoring Causal Effects of Complex Interventions on Networks
Abstract
Complex networks enable formal and extensible representations of real-world systems in a widerange of domains, such as brain networks, transportation networks, power grids, and infectious diseasenetworks. Answering causal questions on complex networks will effectively inform decisionmaking.For instance, in military operations, the commander would be eager to know the potentialcausal effects after an intervention is applied on the complex network, e.g., what will be the overallperformance level (i.e., outcome) of this teamif the unmanned aircraft is dispatched to otherbattlefields (i.e., intervention). Estimating causal effects requires the proper modeling of entities,interventions, and outcomes, while the highly heterogeneous and dynamic nature of complex networkssignificantly increases the difficulty of causal inference. To date, the role of interventions incausal inference on complex networks has been largely underexplored. The current understandingof causal interventions, relevanttheoretical foundations, and computational tools are inadequatefor effectively informing decision-making on complex networks.The proposed research project will investigate theoretical foundations of causal inference on networksand develop new computational tools for monitoring and evaluating complex interventionson networks. Our research plan includes three research tasks. Task 1 aims to infer counterfactualtrajectories of interventions over time and determine the optimal intervention plan. Multipletypes of complex interventions applied on dynamic networks will be considered. Task 2 focuseson mitigating the feature distribution shift and structure distribution shift on networks over time,in order to enhancethe robustness and reliability of causal effects estimation and monitoring ondynamic networks. Task 3 aims to apply causal inference models to new domains, especiallythe non-intervenable networks, in order to enhance the generalizability of models. These threeresearch tasks jointlypresent a unified framework that produce the next-generation reliable andmonitorable causal inference tools for complex networks.The proposed research is highly related to the DoD#s mission in advancing the nation#s artificialintelligence capabilities by laying the groundwork for automated and efficient causal inference oncomplex networks. It will greatly facilitate the development of causality-informed intelligent decisionmaking systems for DoD applications. The proposed causal inference theory and modelsare expected to significantly advance the DoD#s capability in large-scale distributed decision making.Our proposed algorithms and tools can be optimally implemented to extract knowledge andconduct causal reasoning in an open world with large-scale complex networks.This abstract is publicly releasable.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 09, 2024
- Source ID
- N000142412668
Entities
People
- Sheng Li
Organizations
- Office of Naval Research
- United States Navy
- University of Virginia