Academic

Causal Identification from Counterfactual Data: Completeness and Bounding Results

arXiv:2602.23541v1 Announce Type: new Abstract: Previous work establishing completeness results for $\textit{counterfactual identification}$ has been circumscribed to the setting where the input data belongs to observational or interventional distributions (Layers 1 and 2 of Pearl's Causal Hierarchy), since it was generally presumed impossible to obtain data from counterfactual distributions, which belong to Layer 3. However, recent work (Raghavan & Bareinboim, 2025) has formally characterized a family of counterfactual distributions which can be directly estimated via experimental methods - a notion they call $\textit{counterfactual realizabilty}$. This leaves open the question of what $\textit{additional}$ counterfactual quantities now become identifiable, given this new access to (some) Layer 3 data. To answer this question, we develop the CTFIDU+ algorithm for identifying counterfactual queries from an arbitrary set of Layer 3 distributions, and prove that it is complete for this

A
Arvind Raghavan, Elias Bareinboim · · 1 min read · 3 views
#cs.AI #cs.LG

arXiv:2602.23541v1 Announce Type: new Abstract: Previous work establishing completeness results for $\textit{counterfactual identification}$ has been circumscribed to the setting where the input data belongs to observational or interventional distributions (Layers 1 and 2 of Pearl's Causal Hierarchy), since it was generally presumed impossible to obtain data from counterfactual distributions, which belong to Layer 3. However, recent work (Raghavan & Bareinboim, 2025) has formally characterized a family of counterfactual distributions which can be directly estimated via experimental methods - a notion they call $\textit{counterfactual realizabilty}$. This leaves open the question of what $\textit{additional}$ counterfactual quantities now become identifiable, given this new access to (some) Layer 3 data. To answer this question, we develop the CTFIDU+ algorithm for identifying counterfactual queries from an arbitrary set of Layer 3 distributions, and prove that it is complete for this task. Building on this, we establish the theoretical limit of which counterfactuals can be identified from physically realizable distributions, thus implying the $\textit{fundamental limit to exact causal inference in the non-parametric setting}$. Finally, given the impossibility of identifying certain critical types of counterfactuals, we derive novel analytic bounds for such quantities using realizable counterfactual data, and corroborate using simulations that counterfactual data helps tighten the bounds for non-identifiable quantities in practice.

Executive Summary

This article makes significant contributions to the field of causal inference by developing the CTFIDU+ algorithm for identifying counterfactual queries from arbitrary sets of Layer 3 distributions, and establishing a fundamental limit to exact causal inference in the non-parametric setting. The authors leverage recent work on counterfactual realizability to access Layer 3 data, demonstrating that it can be used to identify additional counterfactual quantities. The article also derives novel analytic bounds for non-identifiable quantities using realizable counterfactual data. The findings have important implications for both practical applications and policy-making, as they shed light on the limitations of causal inference and provide a framework for bounding counterfactual quantities.

Key Points

  • ▸ Development of the CTFIDU+ algorithm for identifying counterfactual queries from Layer 3 distributions
  • ▸ Establishment of the fundamental limit to exact causal inference in the non-parametric setting
  • ▸ Derivation of novel analytic bounds for non-identifiable quantities using realizable counterfactual data

Merits

Originality

The article presents novel contributions to the field of causal inference, including the development of a new algorithm and the establishment of a fundamental limit to exact causal inference.

Technical soundness

The authors provide rigorous mathematical derivations and simulations to support their claims, demonstrating a high level of technical soundness.

Demerits

Limited scope

The article is focused on the non-parametric setting, and the results may not generalize to other settings, such as parametric models.

Complexity

The CTFIDU+ algorithm may be computationally intensive and difficult to implement in practice, particularly for large datasets.

Expert Commentary

The article makes significant contributions to the field of causal inference, and the development of the CTFIDU+ algorithm and the establishment of the fundamental limit to exact causal inference in the non-parametric setting are particularly notable. The article's findings have important implications for both practical applications and policy-making, and the novel analytic bounds derived in the article provide a valuable tool for bounding counterfactual quantities. However, the article's limitations, including its focus on the non-parametric setting and the potential complexity of the CTFIDU+ algorithm, should be carefully considered. Overall, the article is a significant contribution to the field of causal inference and has the potential to impact a wide range of applications.

Recommendations

  • ✓ Further research is needed to generalize the article's findings to other settings, such as parametric models.
  • ✓ The CTFIDU+ algorithm should be implemented and tested in practical applications to evaluate its computational efficiency and scalability.

Sources

Related Articles