Argumentative Causal Discovery

TL;DR

This work tackles causal discovery from observational data by introducing Causal ABA, an assumption-based argumentation framework that encodes causal graphs with acyclicity and d-separation as ABA extensions. An ASP-based implementation (ABA-PC) combines independence-test results with domain knowledge to derive DAGs consistent with the data, enabling explainable conflict resolution. Empirical evaluation on bnlearn datasets shows ABA-PC achieving stronger worst-case causal structure accuracy (SID) than several baselines, while highlighting current scalability limitations and potential for integration with other discovery methods. The approach provides a transparent, rule-based pathway to incorporate external evidence and produce faithful causal graphs, with room for extending to latent confounders and improved efficiency.

Abstract

Causal discovery amounts to unearthing causal relationships amongst features in data. It is a crucial companion to causal inference, necessary to build scientific knowledge without resorting to expensive or impossible randomised control trials. In this paper, we explore how reasoning with symbolic representations can support causal discovery. Specifically, we deploy assumption-based argumentation (ABA), a well-established and powerful knowledge representation formalism, in combination with causality theories, to learn graphs which reflect causal dependencies in the data. We prove that our method exhibits desirable properties, notably that, under natural conditions, it can retrieve ground-truth causal graphs. We also conduct experiments with an implementation of our method in answer set programming (ASP) on four datasets from standard benchmarks in causal discovery, showing that our method compares well against established baselines.

Figures (14)

• Figure 1: Overview of the workflow of our Causal ABA algorithm, which combines statistical methods and expert domain knowledge with non-monotonic reasoning and performs argumentative reasoning to output causal graphs consistent with the reported causal relationships.
• Figure 2: Module $\Pi_{\mathit{col}}$
• Figure 3: Module $\Pi_{\mathit{ap}}(\mathscr{p},(v_i)_{i\leq k})$
• Figure 4: Module $\Pi_{\mathit{bp}}(x,y,(\mathscr{p}_i)_{i\leq k})$
• Figure 5: Normalised Structural Interventional Distance for four datasets from the bnlearn repository. Lower is better. Low (resp. High) is the SID for the best (resp. worst) DAG in the estimated CPDAG.

Theorems & Definitions (37)

• Example 1.1
• Definition 2.1
• Definition 2.2
• Definition 2.3
• Definition 2.4
• Definition 2.5
• Example 2.6
• Definition 3.1
• Example 3.2
• Proposition 3.3: restate=PropDAGABA,name=