Causal Inference Isn't Special: Why It's Just Another Prediction Problem

2025-04-06

AI Summary

This paper posits that causal inference is fundamentally a structured prediction problem under distributional shift, where the core challenge lies in extrapolating from labeled source-domain data to unobserved potential outcomes in the target domain—a task inherently afflicted by systematic selection bias due to treatment assignment. To address this, the paper formally establishes, for the first time, that causal assumptions are not stronger but rather more explicit than standard predictive assumptions; it reframes causal estimation as a domain generalization problem with selective labeling, thereby unifying causal inference and predictive modeling within a coherent theoretical framework. Methodologically, it integrates general-purpose machine learning techniques—including importance reweighting, domain adaptation, and counterfactual prediction. The contributions include bridging methodological divides, substantially improving interpretability, cross-domain transferability, and practical usability of causal estimates, while enhancing pedagogical coherence and tool reusability.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

Causal inference is often portrayed as fundamentally distinct from predictive modeling, with its own terminology, goals, and intellectual challenges. But at its core, causal inference is simply a structured instance of prediction under distribution shift. In both cases, we begin with labeled data from a source domain and seek to generalize to a target domain where outcomes are not observed. The key difference is that in causal inference, the labels -- potential outcomes -- are selectively observed based on treatment assignment, introducing bias that must be addressed through assumptions. This perspective reframes causal estimation as a familiar generalization problem and highlights how techniques from predictive modeling, such as reweighting and domain adaptation, apply directly to causal tasks. It also clarifies that causal assumptions are not uniquely strong -- they are simply more explicit. By viewing causal inference through the lens of prediction, we demystify its logic, connect it to familiar tools, and make it more accessible to practitioners and educators alike.

Problem

• Causal inference is a prediction under distribution shift
• Addressing bias in potential outcomes via causal assumptions
• Reframing causal estimation using predictive modeling techniques

Innovation

• Causal inference as prediction under distribution shift
• Apply predictive modeling techniques to causal tasks
• Reframe causal assumptions as explicit generalization constraints