DKCD: Domain Knowledge-Enhanced Causal Discovery from Unstructured Data

2026-07-10 · Source: Computation and Language · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

The Domain Knowledge-enhanced Causal Discovery (DKCD) framework addresses challenges in causal discovery from unstructured data within high-expertise domains like healthcare, finance, and education. Existing large language model (LLM) methods struggle with insufficient identification of latent factors due to a lack of domain-specific knowledge (CH1) and unreliable factor annotation from absent domain-grounded reasoning (CH2), which propagates errors. DKCD introduces three interconnected components: Knowledge Mining, which retrieves relevant domain knowledge; Knowledge-guided Causal Reasoning, which discovers latent causal factors and generates accurate causal clues for annotation; and Causal Structure Discovery, which constructs final causal graphs using a more complete factor set and precise annotations. Experiments on two domain-specific datasets demonstrate that DKCD significantly improves both causal factor identification and causal graph construction.

Key takeaway

For AI Scientists and Research Scientists working on causal discovery from unstructured data in high-expertise domains like healthcare or finance, existing LLM-based methods often fall short due to their inability to identify latent factors and reliably annotate data. You should consider integrating domain-specific knowledge into your causal discovery pipelines, as demonstrated by DKCD, to significantly improve both factor identification and the accuracy of your resulting causal graphs. This approach directly addresses common error propagation issues.

Key insights

DKCD improves causal discovery from unstructured data in high-expertise domains by integrating domain knowledge for latent factor identification and accurate annotation.

Principles

• Domain knowledge is vital for latent factor discovery.
• Domain-grounded reasoning prevents annotation errors.

Method

DKCD's method involves mining domain knowledge, guiding causal reasoning to discover latent factors and generate annotation clues, then constructing causal graphs from complete, accurate data.

In practice

• Apply DKCD in healthcare for disease causality.
• Use DKCD for financial risk factor identification.
• Enhance educational outcome analysis with DKCD.

Topics

• Causal Discovery
• Unstructured Data Analysis
• Domain Knowledge Integration
• Latent Factor Identification
• Causal Graph Construction
• Large Language Models

Best for: AI Scientist, Research Scientist

Related on AIssential

• DKCD: Domain Knowledge-Enhanced Causal Discovery from Unstructured Data — Integrating domain knowledge graphs with LLMs significantly enhances causal discovery from unstructured data in high-expertise fields.
• Unsupervised Causal Abstractions Discovery — This work proposes an unsupervised method to discover high-level causal abstractions directly from low-level observational data.
• Tensor-based second-order causal discovery — TSCD uses second-order statistics from observational and interventional data to efficiently discover causal graphs and functions, scaling to hundreds of variables.
• A Survey on Federated Causal Discovery and Inference — Federated Causal Discovery and Inference are complementary stages in a unified pipeline for privacy-preserving causal reasoning.
• Latent Confounded Causal Discovery via Lie Bracket Geometry — Latent causal structure can be inferred directly from the geometry of intervention-induced flows and Lie bracket failures.
• CDFM: Towards a General-Purpose Causal Discovery Foundation Model — CDFM unifies causal discovery by learning generalizable representations from diverse mechanisms for zero-shot structural inference.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.