DODO: Causal Structure Learning with Budgeted Interventions

2025/10/09 by Matteo Gregorini, Gregorini, Matteo, Chiara Boldrini +3
Computer Science · #Artificial Intelligence (cs.AI) #Bayesian Modeling and Causal Inference #Cognitive Science and Mapping #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences

paper · pdf · doi:10.48550/arxiv.2510.08207

openalex publication_date 2025/10/09 · openalex created_date 2025/10/11 · openalex updated_date 2026/07/28

Abstract

Artificial Intelligence has achieved remarkable advancements in recent years, yet much of its progress relies on identifying increasingly complex correlations. Enabling causality awareness in AI has the potential to enhance its performance by enabling a deeper understanding of the underlying mechanisms of the environment. In this paper, we introduce DODO, an algorithm defining how an Agent can autonomously learn the causal structure of its environment through repeated interventions. We assume a scenario where an Agent interacts with a world governed by a causal Directed Acyclic Graph (DAG), which dictates the system's dynamics but remains hidden from the Agent. The Agent's task is to accurately infer the causal DAG, even in the presence of noise. To achieve this, the Agent performs interventions, leveraging causal inference techniques to analyze the statistical significance of observed changes. Results show better performance for DODO, compared to observational approaches, in all but the most limited resource conditions. DODO is often able to reconstruct with as low as zero errors the structure of the causal graph. In the most challenging configuration, DODO outperforms the best baseline by +0.25 F1 points.

Citations

• Voyager: An Open-Ended Embodied Agent with Large Language Models
• A Survey on Causal Reinforcement Learning
• Learning Neural Causal Models with Active Interventions
• Differentiable Causal Discovery from Interventional Data
• Causality for Machine Learning
• Causal Discovery with Reinforcement Learning
• Review of Causal Discovery Methods Based on Graphical Models
• Explainable Reinforcement Learning Through a Causal Lens
• Causal Reasoning from Meta-reinforcement Learning
• DAGs with NO TEARS: Continuous Optimization for Structure Learning
• Causal Bandits: Learning Good Interventions via Causal Inference
• Learning Causal Graphs with Small Interventions
• Distinguishing cause from effect using observational data: methods and benchmarks
• Identifiability of Gaussian structural equation models with equal error variances
• Kernel-based Conditional Independence Test and Application in Causal Discovery
• On the completeness of orientation rules for causal discovery in the presence of latent confounders and selection bias
• Estimating high-dimensional directed acyclic graphs with the PC-algorithm
• On the Number of Experiments Sufficient and in the Worst Case Necessary to Identify All Causal Relations Among N Variables
• Learning Gaussian Networks
• Approximating probabilistic inference in Bayesian belief networks is NP-hard
• An Algorithm for Fast Recovery of Sparse Causal Graphs

Related

• From Statistical to Causal Learning
• Causal Discovery and Prediction: Methods and Algorithms
• BISCUIT: Causal Representation Learning from Binary Interactions
• Better Simulations for Validating Causal Discovery with the DAG-Adaptation of the Onion Method
• A Meta-Reinforcement Learning Algorithm for Causal Discovery