# Causal Policy Learning in Reinforcement Learning: Backdoor-Adjusted Soft Actor-Critic

arXiv: https://arxiv.org/abs/2506.05445 · HTML: https://mlsanity.com/paper/2506.05445/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Deep RL algorithms and representations, one of 15 directions in Deep reinforcement learning methods.
- Builds on 24 of its 47 references that are in the map.
- First posted on arXiv 5 Jun 2025.

## Placed in

- Deep reinforcement learning methods / [Deep RL algorithms and representations](https://mlsanity.com/deeprl/deep-rl-algorithms-and-representations/)

## Builds on (most cited first, 24 of 24)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor](https://mlsanity.com/paper/1801.01290/) (2018)
- [Continuous control with deep reinforcement learning](https://mlsanity.com/paper/1509.02971/) (2015)
- [Addressing Function Approximation Error in Actor-Critic Methods](https://mlsanity.com/paper/1802.09477/) (2018)
- [OpenAI Gym](https://mlsanity.com/paper/1606.01540/) (2016)
- [Causality for Machine Learning](https://mlsanity.com/paper/1911.10500/) (2019)
- [Counterfactual Off-Policy Evaluation with Gumbel-Max Structural Causal Models](https://mlsanity.com/paper/1905.05824/) (2019)
- [Federated Reinforcement Learning with Environment Heterogeneity](https://mlsanity.com/paper/2204.02634/) (2022)
- [Off-policy Policy Evaluation For Sequential Decisions Under Unobserved Confounding](https://mlsanity.com/paper/2003.05623/) (2020)
- [Sample-Efficient Reinforcement Learning via Counterfactual-Based Data Augmentation](https://mlsanity.com/paper/2012.09092/) (2020)
- [Causal Curiosity: RL Agents Discovering Self-supervised Experiments for Causal Representation Learning](https://mlsanity.com/paper/2010.03110/) (2020)
- [Causal Imitation Learning with Unobserved Confounders](https://mlsanity.com/paper/2208.06267/) (2022)
- [Provably Efficient Causal Reinforcement Learning with Confounded Observational Data](https://mlsanity.com/paper/2006.12311/) (2020)
- [Budgeted Experiment Design for Causal Structure Learning](https://mlsanity.com/paper/1709.03625/) (2017)
- [A Minimax Learning Approach to Off-Policy Evaluation in Confounded Partially Observable Markov Decision Processes](https://mlsanity.com/paper/2111.06784/) (2021)
- [Causal Imitation Learning under Temporally Correlated Noise](https://mlsanity.com/paper/2202.01312/) (2022)
- [Sequential Causal Imitation Learning with Unobserved Confounders](https://mlsanity.com/paper/2208.06276/) (2022)
- [Federated Causal Inference in Heterogeneous Observational Data](https://mlsanity.com/paper/2107.11732/) (2021)
- [Highly Efficient Self-Adaptive Reward Shaping for Reinforcement Learning](https://mlsanity.com/paper/2408.03029/) (2024)
- [An Adaptive Kernel Approach to Federated Learning of Heterogeneous Causal Effects](https://mlsanity.com/paper/2301.00346/) (2023)
- [ACE : Off-Policy Actor-Critic with Causality-Aware Entropy Regularization](https://mlsanity.com/paper/2402.14528/) (2024)
- [Robust Policy Optimization in Deep Reinforcement Learning](https://mlsanity.com/paper/2212.07536/) (2022)
- [Delphic Offline Reinforcement Learning under Nonidentifiable Hidden Confounding](https://mlsanity.com/paper/2306.01157/) (2023)
- [Two-way Deconfounder for Off-policy Evaluation in Causal Reinforcement Learning](https://mlsanity.com/paper/2412.05783/) (2024)

## Abstract (arXiv)

Hidden confounders that influence both states and actions can bias policy learning in reinforcement learning (RL), leading to suboptimal or non-generalizable behavior. Most RL algorithms ignore this issue, learning policies from observational trajectories based solely on statistical associations rather than causal effects. We propose DoSAC (Do-Calculus Soft Actor-Critic with Backdoor Adjustment), a principled extension of the SAC algorithm that corrects for hidden confounding via causal intervention estimation. DoSAC estimates the interventional policy π(a | do(s)) using the backdoor criterion, without requiring access to true confounders or causal labels. To achieve this, we introduce a learnable Backdoor Reconstructor that infers pseudo-past variables (previous state and action) from the current state to enable backdoor adjustment from observational data. This module is integrated into a soft actor-critic framework to compute both the interventional policy and its entropy. Empirical results on continuous control benchmarks show that DoSAC outperforms baselines under confounded settings, with improved robustness, generalization, and policy reliability.