Abstract
Title: Knowledge-Augmented Causal Discovery: Integrating LLM-Based Knowledge Systems with Data-Driven Structure Learning
Abstract: Causal discovery from data is limited by the structure of available datasets (finite samples, measurement error, unmeasured confounding, incomplete variable coverage), by the absence of causal logic in purely data-driven algorithms, and by the lack of external domain knowledge such as pathophysiology and temporal ordering that experts routinely use to constrain causal claims. We will discuss Knowledge-Augmented Causal Discovery (KACD) that combines data-driven structural learning with LLM-based knowledge systems to address these limitations. Knowledge derived from the literature and LLM reasoning informs the discovery process, for example by constraining the search space and supporting edge orientation. We apply KACD to chronic low back pain (cLBP), a heterogeneous, multifactorial condition in which biological, psychological, and social factors interact and the causal structure remains contested. We examine whether knowledge augmentation improves the plausibility and stability of the learned graphs relative to data-driven discovery alone, and where the knowledge system helps.
Speakers
Moderator
About the Speakers
Abel Torres-Espin is an Assistant Professor in the School of Public Health Sciences at the University of Waterloo. His research focuses on the application of data science methods to health research, with a particular interest in conceptualizing health and disease as complex, dynamic processes that span multiple domains, including biological, medical, psychological, social, and technological factors, and multiple levels, from molecular to individual and population health. His work applies health data science, artificial intelligence (AI), and machine learning approaches to integrate diverse data resources, answer critical health-related questions, and inform evidence-based practice and policy. With increasing health and health-care challenges, the growing availability of digital health data provides unprecedented opportunities to better understand this complexity and advance precision health and evidence-informed decision-making. Dr. Torres-Espin's research has focused on complex and heterogeneous populations, including individuals with spinal cord injury, traumatic brain injury, dementia, and chronic low back pain, conditions that result in substantial individual and societal burdens. He is also interested in analytical and computational methods, multivariate statistics, research reproducibility and replicability, and the generation of real-world evidence to support health care and public health decision-making. See Profile
Paul E. Anderson is a Professor of Computer Science and Software Engineering at California Polytechnic State University, San Luis Obispo, where he leads the Data Science Research Group. His research focuses on advancing computational methods for scientific discovery, with expertise in artificial intelligence, bioinformatics, data science, and biomedical informatics. His work spans genomics, proteomics, metabolomics, ranking algorithms, and algorithmic fairness, with applications in toxicology, marine biology, cancer informatics, and other data-intensive scientific domains. He is particularly interested in the foundations of data science and the development of explainable analytical methods that support impactful research across disciplines. In addition to his research, Dr. Anderson serves as a Faculty Fellow with Cal Poly's Center for Innovation and Entrepreneurship, where he mentors students and promotes innovation within engineering education. He earned his B.S. in Computer Engineering, M.S. in Computer Science, and Ph.D. in Computer Science and Engineering from Wright State University and has held positions in academia and government research, including the Air Force Research Laboratory.
About the Moderator
Joel Dubin, University of Waterloo
About the NISS-CANSSI Collaborative Data Science Web Series:
The NISS-CANSSI Collaborative Data Science initiative that the National Institute of Statistical Sciences (NISS) in collaboration with the Canadian Statistical Sciences Institute (CANSSI) brings together experts from various fields to tackle complex data challenges through interdisciplinary teamwork and innovative methodologies.
Goals of the Initiative
The goal is to foster progress in:
- Developing new ideas for experimental and observational data-driven learning and discovery that address key questions at the cutting edge of science and scientific deduction;
- Quantifying and summarizing uncertainty in data-driven theories, as well as complex Data Science models, algorithms, and workflows; and
- Establishing new practices for scientific reproducibility and replicability through Data Science.
Learn more: NISS-CANSSI Collaborative Data Science
Event Type
- NISS Hosted
Host
Cost
Location
Policy
