III: Small: pCAR: Discovering and Leveraging Plausibly Causal (p-causal) Relationships to Understand Complex Dynamic Systems

Arizona State University, Scottsdale AZ

Investigators

Abstract

Today, successfully tackling many urgent challenges in socio-economically critical domains (such as sustainability, public health, and biology) requires obtaining a deeper understanding of causal relationships and interactions among a diverse spectrum of entities from high-dimensional data. For example, the emergence and propagation of an epidemic involves a causally complex interplay of entities, including individuals and their social-interactions, physical short-range and long-range networks of mobility, and intervention decisions (such as school closures or restrictions on mobility) by decision makers. With the aim of filling this important hole and enabling applications and services with significant economic and public health impact, the project carries out research on causally-aware data-driven science and engineering, including data-supported causal analysis. The project also contributes to workforce growth, provides an excellent context for doctoral, master's, and undergraduate level research and teaching. Future researchers that have the skills in both data analysis and management are being trained and students are being introduced to career pathways through their participation in research, publications, and partnerships with domain experts. In this project, the research team hypothesizes that data analysis provides opportunities for identifying relationships that can potentially be causal. They further hypothesize that data can be used for strengthening and weakening causal assumptions, and for pruning relationships that are certainly not causal. To validate and leverage these hypotheses, they introduce the novel concepts of 'plausible causality' (p-causality) and 'plausibly causal (p-causal) relationships' and they develop techniques to (i) discover p-causal interactions and relationships, (ii) maintain these p-casual relationships in systems where causality itself evolves over time, and (iii) use discovered p-causal relationships to support efficient and effective data analytics to study complex, dynamic systems. In particular, they develop new models to capture context-sensitive plausibly causal (p-causal) relationships in complex, dynamic systems and design novel and scalable causally-aware data analysis algorithms that leverage known or hypothesized p-causal relationships among entities within different and potentially evolving contexts to deal with data sparsity, imprecision, and noise. 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.