The paper proposes a novel method for causal inference using invariant prediction, which leverages the invariance property of causal models under different experimental settings or interventions. The authors introduce a framework where the conditional distribution of the target variable given its direct causal predictors remains unchanged across various environments. This property is exploited to construct confidence intervals for causal relationships and identify causal predictors. The method is applicable to both observational and interventional data, and it does not require prior knowledge of the causal ordering of variables. The paper provides theoretical guarantees for the method's validity and discusses its robustness to model misspecification. Two specific methods for estimating causal predictors and their associated coefficients are presented, along with empirical results from various datasets, including gene perturbation experiments and educational studies. The method is implemented in the R package InvariantCausalPrediction.The paper proposes a novel method for causal inference using invariant prediction, which leverages the invariance property of causal models under different experimental settings or interventions. The authors introduce a framework where the conditional distribution of the target variable given its direct causal predictors remains unchanged across various environments. This property is exploited to construct confidence intervals for causal relationships and identify causal predictors. The method is applicable to both observational and interventional data, and it does not require prior knowledge of the causal ordering of variables. The paper provides theoretical guarantees for the method's validity and discusses its robustness to model misspecification. Two specific methods for estimating causal predictors and their associated coefficients are presented, along with empirical results from various datasets, including gene perturbation experiments and educational studies. The method is implemented in the R package InvariantCausalPrediction.