CausalForge: A Formally Grounded, Self-Improving Agentic Framework for Automated Research in Causal Inference

Researchers introduced CausalForge, an agentic framework for automated causal-inference research built on Lean-based formal verification.

It combines a machine-checked causal inference library with an autonomous pipeline that generates topics, conjectures, proofs, and research artifacts.

A separate audit step checks whether formal theorems match the intended informal claims, helping catch mismatches in meaning.

The approach addresses a key weakness in AI-assisted research: producing plausible results is easier than proving they are correct and meaningful.