FinCARE: Financial Causal Analysis with Reasoning and Evidence

Decision-making tools for financial services often depend on correlation-based analysis and heuristics. While these methods can pinpoint relationships between variables, they often fail to dig deeper to explain why those relationships exist— or if they hold true at all. This lack of cause-and-effect understanding can lead to misguided investments, weak risk management, and other costly consequences that regulated industries like financial services simply cannot afford to face. 

But that’s about to change. Domyn’s research team is introducing the FinCARE framework, which brings scientific rigor and causal reasoning to financial analysis. Led by Bhaskarjit Sarmah, Stefano Pasquali, Abhinav Arun, and Alejandro Zuñiga, FinCARE is a hybrid framework that combines statistical causal discovery algorithms with financial Knowledge Graphs (KGs) extracted from SEC 10-K filings, as well as the reasoning capabilities of Large Language Models (LLMs). This synergy provides a more robust understanding of causal mechanisms in finance, giving FinCARE an edge that traditional methods —or even advanced AI models — can’t match. 

FinCARE’s early results exceeded expectations. All tested algorithms observed a substantial improvement in terms of performance. More importantly, it delivered reliable counterfactual predictions, achieving near-zero error and perfect directional accuracy for intervention effects. In other words, FinCARE can simulate “what-if” scenarios — such as regulatory changes or market shocks — before they take place, providing a real competitive advantage for proactive risk management and strategic decision-making. 

These findings only scratch the surface of how FinCARE can help foster trustworthy and effective AI solutions for financial services.