DAGForge: Browser-Based Tool for Auditable Causal DAG Authoring
DAGForge, a novel browser-based tool, has been created by researchers to facilitate the construction of causal directed acyclic graphs (DAGs) that are auditable and linked to evidence. This system simplifies the traditionally manual task of correlating study variables with biomedical literature, assessing uncertain causal assertions, and maintaining provenance for expert evaluation. It generates a reproducible snapshot of literature from free-text study descriptions, employs a reasoning module based on LLM to produce structured causal judgments supported by direct evidence excerpts, and organizes these into a constraint-validated graph. Each edge proposed includes confidence metrics, provenance information, and a rationale for review. The interface allows for study specification, progress tracking, evidence evaluation, graph comparisons, adjustment-set calculations, and data export. The system's performance was assessed against compact benchmarks, though detailed results are not included in the abstract. This research is available on arXiv with the identifier 2607.21859.
Key facts
- DAGForge is a browser-based system for authoring causal DAGs.
- It creates auditable, evidence-linked artifacts.
- It uses an LLM-based reasoning module for pairwise causal judgments.
- Each edge includes confidence estimates, provenance, and rationale.
- The interface supports study specification, progress monitoring, evidence review, graph comparison, adjustment-set computation, and export.
- The system was evaluated against compact benchmarks.
- The paper is available on arXiv with ID 2607.21859.
- The tool is designed for biomedical causal analysis.
Entities
Institutions
- arXiv