Causal Discovery as Dialectical Aggregation: A Quantitative Argumentation Framework

2026-04-26

AI Summary

Constraint-based causal discovery methods are prone to structural errors under limited sample sizes due to misjudgments in conditional independence tests. This work proposes a novel framework grounded in quantitative argumentation, treating conditional independence test outcomes as defeasible, strength-weighted arguments. Conflicting evidence is aggregated through a connectivity-guided witness propagation mechanism, and the acceptability of candidate edges is determined via fixed-point labeling. By introducing dialectical aggregation into causal discovery for the first time, the approach replaces rigid constraints with flexible argumentation, substantially enhancing structural robustness. Evaluated on standard Bayesian network benchmarks, the method outperforms classical constraint-based, hybrid, and existing argumentation-based approaches under noisy or inconsistent conditions, demonstrating superior performance in both structural consistency and interventional reliability.

Technology Category

• Reasoning under Uncertainty: Causality
• Knowledge Representation and Reasoning: Argumentation
• Constraint Satisfaction and Optimization: Other Foundations of Constraint Satisfaction

Application Category

• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Constraint-based causal discovery is brittle in finite-sample regimes because erroneous conditional-independence (CI) decisions can cascade into substantial structural errors. We propose Quantitative Argumentation for Causal Discovery (QACD), a semantics-driven framework that represents CI outcomes as graded, defeasible arguments rather than irreversible constraints. QACD maps statistical test outcomes to argument strengths and aggregates conflicting evidence through connectivity-mediated witness propagation, producing a fixed-point acceptability labeling over candidate adjacencies. Experiments on standard benchmark Bayesian networks suggest that QACD improves structural coherence and interventional reliability in several noisy or inconsistent CI regimes, while remaining competitive with classical constraint-based, hybrid, and prior argumentation-based baselines.

Problem

• causal discovery
• conditional independence
• finite-sample regimes
• structural errors
• constraint-based methods

Innovation

• causal discovery
• quantitative argumentation
• conditional independence
• defeasible reasoning
• witness propagation