CausalSteward: An Agentic Divide-Conquer-Combine Copilot for Causal Discovery
What this work does
Partitions high-dimensional variables, combines LLM/human prior knowledge with data-driven discovery in local graphs, and merges results through hypothesis/critic agents.
Connection to discovery
Exposes a testable boundary between domain-knowledge proposals, statistical tests and final graph commitments in a harness.
Evidence
Inspected how proposed graphs constrain a discovery algorithm and how cross-partition edges are added during merging.
Scope and limitations
Incorrect priors and partitioning can bias graphs; identifiability claims depend on assumptions and available evidence. The selected review did not independently check the theoretical proofs or reproduce experiments.
THE RESEARCH CONVERSATION
Discussion & reactions
Share a question, an insight, or a replication experience. Sign in with GitHub to join the conversation.