From Identifiable Causal Representations to Controllable Counterfactual Generation: A Survey on Causal Generative Modeling
Aneesh Komanduri, Xintao Wu, Yongkai Wu, Feng Chen
TL;DR
This survey presents a comprehensive overview of causal generative modeling, focusing on identifiable causal representation learning (CRL) and controllable counterfactual generation (CCG). It articulates a Pearl-based causal hierarchy and reviews methodological families across observational, interventional, and counterfactual data regimes, including VAE-, flow-, and diffusion-based approaches. Key contributions include a taxonomy of CRL methods (e.g., CausalVAE, DEAR, SCM-VAE, ICM-VAE) with identifiability guarantees under auxiliary signals, multi-domain data, and temporal settings, plus a taxonomy of CCG techniques (GAN-, flow-, and diffusion-based) enabling counterfactual generation with tractable inference. The paper also surveys datasets, evaluation metrics, and real-world applications in fairness, privacy, robustness, and precision medicine, and outlines open problems and future directions for scalable, verifiable CRL/CCG with weaker supervision and stronger benchmarks.
Abstract
Deep generative models have shown tremendous capability in data density estimation and data generation from finite samples. While these models have shown impressive performance by learning correlations among features in the data, some fundamental shortcomings are their lack of explainability, tendency to induce spurious correlations, and poor out-of-distribution extrapolation. To remedy such challenges, recent work has proposed a shift toward causal generative models. Causal models offer several beneficial properties to deep generative models, such as distribution shift robustness, fairness, and interpretability. Structural causal models (SCMs) describe data-generating processes and model complex causal relationships and mechanisms among variables in a system. Thus, SCMs can naturally be combined with deep generative models. We provide a technical survey on causal generative modeling categorized into causal representation learning and controllable counterfactual generation methods. We focus on fundamental theory, methodology, drawbacks, datasets, and metrics. Then, we cover applications of causal generative models in fairness, privacy, out-of-distribution generalization, precision medicine, and biological sciences. Lastly, we discuss open problems and fruitful research directions for future work in the field.