Cyclic Counterfactuals under Shift-Scale Interventions

2025-10-28

AI Summary

Traditional counterfactual reasoning relies on acyclic structural causal models (SCMs), limiting its applicability to real-world systems with feedback loops—e.g., biological regulatory networks. This work extends counterfactual inference to **general cyclic SCMs**, focusing on **shift-scale soft interventions**, i.e., differentiable translations and scalings of mechanism functions. We propose a computational framework based on implicit function differentiation and differentiable optimization, enabling stable solution of nonlinear equation systems induced by mechanism transformations. Our method yields differentiable and consistent estimation of counterfactual distributions in cyclic systems. Experiments demonstrate its effectiveness and robustness on both synthetic cyclic SCMs and real biological pathways. To our knowledge, this is the first theoretically sound and computationally tractable counterfactual analysis tool for complex systems exhibiting strong feedback dynamics.

Technology Category

Reasoning under Uncertainty: CausalityMachine Learning: Causal LearningCognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingUser Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systemsGraph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

Most counterfactual inference frameworks traditionally assume acyclic structural causal models (SCMs), i.e. directed acyclic graphs (DAGs). However, many real-world systems (e.g. biological systems) contain feedback loops or cyclic dependencies that violate acyclicity. In this work, we study counterfactual inference in cyclic SCMs under shift-scale interventions, i.e., soft, policy-style changes that rescale and/or shift a variable's mechanism.

Problem

• Extending counterfactual inference to cyclic structural causal models
• Addressing feedback loops in real-world systems like biology
• Studying shift-scale interventions as soft policy-style changes

Innovation

• Cyclic structural causal models for counterfactual inference
• Shift-scale interventions modifying variable mechanisms
• Handling feedback loops in real-world biological systems