Reproduce “CausalKG: Causal Knowledge Graph Explainability using interventional and counterfactual reasoning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Utkarshani Jaimini, Amit Sheth
Researchers from the Artificial Intelligence Institute (AIISC) at the University of South Carolina developed CausalKG, a framework for representing causal relationships in Knowledge Graphs to enhance AI explainability. This framework enables interventional and counterfactual reasoning by leveraging Causal Bayesian Networks and RDF*, effectively demonstrating its application in autonomous driving scenarios for understanding complex accident causes and hypothetical situations.
Key insights
- A blend of Bayesian causality with advanced Knowledge Graph representation (RDF*) allows for robust encoding and reasoning with complex, multi-entity causal relationships.
- Domain explainability can be achieved by enabling AI systems to answer "what if" and "what would have happened" questions through explicitly modeled causal mechanisms, going beyond purely associational reasoning.
- RDF* offers a more intuitive and concise method for representing hyper-relational data within KGs, particularly for annotating causal links directly with quantitative effect metadata.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.