Machine Learning for Causal Inference

Traditional spatiotemporal data analysis often relies on predictive models that overlook causal relationships, making it difficult to identify true drivers and formulate effective interventions. To bridge this gap, we review causal machine learning (CML) techniques for spatiotemporal data, aiming to provide robust insights into their unique advantages. Our literature review reveals that fewer than 1% of studies in major databases explicitly integrate CML with spatiotemporal analysis. After rigorous screening, we analyze 51 relevant papers, categorizing their contributions into four key areas (totaling 62 methodological approaches due to multi-category papers): 1) causal effect discovery and estimation (32 approaches), 2) prediction accuracy enhancement (19), 3) pattern recognition limitations (10), and 4) interpretability (1). This distribution highlights a critical research gap, particularly in interpretability and comprehensive frameworks. We further examine unique challenges in spatiotemporal data, such as spatial autocorrelation and temporal dependencies, that complicate causal inference but also present opportunities for innovation. Promising approaches include the synergy of spatiotemporal Granger causality and structural equation modeling with spatial lags, which capture complex interdependencies while preserving interpretability. Future directions include developing interpretable causal models, advancing real-time causal inference in dynamic environments, and addressing computational challenges (scalability, efficiency, and complexityinterpretability trade-offs). We also discuss ethical considerations, such as bias mitigation in causal discovery and societal implications of spatiotemporal causal inference. By synthesizing challenges and opportunities, this work advances the application of CML in spatiotemporal analysis, with implications for climate science, economics, epidemiology, and urban planning. INDEX TERMS Causal machine learning, spatiotemporal data analysis, synergy methods, ethics.

“Causal Computation Theory” introduces a comprehensive framework for structuring computation around causal reasoning, enabling intelligent systems to move beyond pattern recognition into structured, adaptive decision-making under uncertainty.
This volume covers:
✅ Foundations of causal computation, causal graphs, and structural causal models.
✅ Intervention and counterfactual reasoning within computational systems.
✅ Entropy flows as computational drivers under causal constraints.
✅ Architectures and algorithms for building causal machines.
✅ Future directions in scalable causal learning, hybrid causal-deep models, and AI alignment.
Positioned at the intersection of information theory, computational systems, and causal inference, this work offers researchers, system designers, and advanced learners a structured pathway to integrate causal reasoning into artificial intelligence, robotics, and complex systems analysis.
Computation structures entropy into information. Causal computation aligns this structure with the fabric of the world.

Effective customer relationship management (CRM) techniques are essential in today's business environments for companies looking to maximize client interactions and increase revenue. This paper addresses customer churn, a significant issue that firms in various industries confront, by providing a thorough examination of how to improve CRM through the integration of AI and ML technologies. The importance of CRM and how AI and ML are transforming customer interactions, customization, and operational efficiency are covered in detail in the first section of the study. An investigation of a case study is carried out to see how well different machine learning models predict customer attrition in CRM systems. The following models are evaluated: GaussianNB, Artificial Neural Networks (ANN), K-Neighbors Classifier, Support Vector Classifier (SVC), Decision Tree, Random Forest, and Logistic Regression. With an accuracy of 92.5%, Random Forest Classifier is found to be the most successful model in the study; Decision Tree Classifier is next closest at 89.8%. It also looks at how important features engineering, data preparation, model selection, training, validation, deployment, and performance tracking are for AI-driven CRM systems. Data features, outlier identification, linear correlations, and model accuracy evaluation are all made possible by visualizations such as histograms, box plots, scatterplots, and performance metrics for classification models. The results highlight how crucial data quality, algorithm selection, and continuous model monitoring are to the success of CRM projects. By utilizing AI and ML technology, this research propels CRM approaches forward, enabling firms to anticipate consumer behaviors, tailor interactions, and cultivate enduring customer connections in highly competitive market conditions.

This is one of the challenges that new and fast-growing econometric literature is beginning to tackle in addressing causal inference problems with machine learning methods. Yet, empirical economics still has not really made use of the strengths of these modern approaches. Here, we revisit groundbreaking empirical work through the perspective of causal machine learning methods to connect econometric theory with applied economics. In particular, we will cover double machine learning, causal forests, and more general machine learning methodologies, both in the setting of average treatment effects and heterogeneous treatment effects. We demonstrate the application of these methods in diverse settings and discuss their significance and additional benefits relative to classical approaches that were utilized in the original studies.

Customer churn describes terminating a relationship with a business or reducing customer engagement over a specific period. Customer acquisition cost can be five to six times that of customer retention, hence investing in customers with churn risk is wise. Causal analysis of the churn model can predict whether a customer will churn in the foreseeable future and identify effects and possible causes for churn. In general, this study presents a conceptual framework to discover the confounding features that correlate with independent variables and are causally related to those dependent variables that impact churn. We combine different algorithms including the SMOTE, ensemble ANN, and Bayesian networks to address churn prediction problems on a massive and high-dimensional finance data that is usually generated in financial institutions due to employing interval-based features used in Customer Relationship Management systems. The effects of the curse and blessing of dimensionality assess...

We propose counterfactual reasoning through probabilistic
logic twin networks (PLTNs) to prevent collisions in self-driving cars.
The basis of a PLTNs is a causal Bayesian network (cBN ) partially
learned from simulated self-driving car data and synthetic data. The
cBN includes the lane of the self-driving car, the presence of up to 4
surrounding vehicles, an indicator for potential collisions, and 6 driving
actions. Counterfactual queries through the PLTNs intervene with alternative actions to identify which minimizes the probability of a collision.
For evaluation, three cBNs are learned with 1%, 50%, and 100% of a training dataset. For querying, 120 state-action examples labeled as leading
to a crash are selected randomly. Each one is associated with six possible
interventions. The probability of a collision is then queried from PLTNs,
provided that a potential collision has been warned, and the current state
and action are known. Results show that all intervened actions minimizing the probability of a crash does not lead to a car crash, suggesting the
effectiveness of this approach in developing collision prevention schemes
for self-driving cars. To the best of our knowledge, this is the first application of PLTNs for counterfactual reasoning in autonomous vehicles

Customer churn describes terminating a relationship with a business or reducing customer engagement over a specific period. Customer acquisition cost can be five to six times that of customer retention, hence investing in customers with churn risk is wise. Causal analysis of the churn model can predict whether a customer will churn in the foreseeable future and identify effects and possible causes for churn. In general, this study presents a conceptual framework to discover the confounding features that correlate with independent variables and are causally related to those dependent variables that impact churn. We combine different algorithms including the SMOTE, ensemble ANN, and Bayesian networks to address churn prediction problems on a massive and high-dimensional finance data that is usually generated in financial institutions due to employing interval-based features used in Customer Relationship Management systems. The effects of the curse and blessing of dimensionality assessed by utilising the Recursive Feature Elimination method to overcome the high dimension feature space problem. Moreover, a causal discovery performed to find possible interpretation methods to describe cause probabilities that lead to customer churn. Evaluation metrics on validation data confirm the random forest and our ensemble ANN model, with %86 accuracy, outperformed other approaches. Causal analysis results confirm that some independent causal variables representing the level of super guarantee contribution, account growth, and account balance amount were identified as confounding variables that cause customer churn with a high degree of belief. This article provides a real-world customer churn analysis from current status inference to future directions in local superannuation funds.

We apply causal machine learning algorithms to assess the causal effect of a marketing intervention, namely a coupon campaign, on the sales of a retailer. Besides assessing the average impacts of different types of coupons, we also investigate the heterogeneity of causal effects across different subgroups of customers, e.g., between clients with relatively high vs. low prior purchases. Finally, we use optimal policy learning to determine (in a data-driven way) which customer groups should be targeted by the coupon campaign in order to maximize the marketing intervention's effectiveness in terms of sales. We find that only two out of the five coupon categories examined, namely coupons applicable to the product categories of drugstore items and other food, have a statistically significant positive effect on retailer sales. The assessment of group average treatment effects reveals substantial differences in the impact of coupon provision across customer groups, particularly across customer groups as defined by prior purchases at the store, with drugstore coupons being particularly effective among customers with high prior purchases and other food coupons among customers with low prior purchases. Our study provides a use case for the application of causal machine learning in business analytics to evaluate the causal impact of specific firm policies (like marketing campaigns) for decision support.

At the heart of causal structure learning from observational data lies a deceivingly simple question: given two statistically dependent random variables, which one has a causal effect on the other? This is impossible to answer using statistical dependence testing alone and requires that we make additional assumptions. We propose several fast and simple criteria for distinguishing cause and effect in pairs of discrete or continuous random variables. The intuition behind them is that predicting the effect variable using the cause variable should be `simpler' than the reverse -- different notions of `simplicity' giving rise to different criteria. We demonstrate the accuracy of the criteria on synthetic data generated under a broad family of causal mechanisms and types of noise.

At the heart of causal structure learning from observational data lies a deceivingly simple question: given two statistically dependent random variables, which one has a causal effect on the other? This is impossible to answer using statistical dependence testing alone and requires that we make additional assumptions. We propose fast and simple criteria for distinguishing cause and effect in pairs of discrete or continuous random variables. The intuition behind them is that predicting the effect variable using the cause variable should be 'simpler' than the reverse different notions of 'simplicity' giving rise to different criteria. We demonstrate the accuracy of the criteria on synthetic data generated under a broad family of causal mechanisms and types of noise.

At the heart of causal structure learning from observational data lies a deceivingly simple question: given two statistically dependent random variables, which one has a causal effect on the other? This is impossible to answer using statistical dependence testing alone and requires that we make additional assumptions. We propose several fast and simple criteria for distinguishing cause and effect in pairs of discrete or continuous random variables. The intuition behind them is that predicting the effect variable using the cause variable should be `simpler' than the reverse -- different notions of `simplicity' giving rise to different criteria. We demonstrate the accuracy of the criteria on synthetic data generated under a broad family of causal mechanisms and types of noise.

We consider evaluating the causal effects of dynamic treatments, i.e. of multiple treatment sequences in various periods, based on double machine learning to control for observed, time-varying covariates in a data-driven way under a selection-on-observables assumption. To this end, we make use of so-called Neyman-orthogonal score functions, which imply the robustness of treatment effect estimation to moderate (local) misspecifications of the dynamic outcome and treatment models. This robustness property permits approximating outcome and treatment models by double machine learning even under high dimensional covariates and is combined with data splitting to prevent overfitting. In addition to effect estimation for the total population, we consider weighted estimation that permits assessing dynamic treatment effects in specific subgroups, e.g. among those treated in the first treatment period. We demonstrate that the estimators are asymptotically normal and √ n-consistent under specific regularity conditions and investigate their finite sample properties in a simulation study. Finally, we apply the methods to the Job Corps study in order to assess different sequences of training programs under a large set of covariates.

At the heart of causal structure learning from observational data lies a deceivingly simple question: given two statistically dependent random variables, which one has a causal effect on the other? This is impossible to answer using statistical dependence testing alone and requires that we make additional assumptions. We propose fast and simple criteria for distinguishing cause and effect in pairs of discrete or continuous random variables. The intuition behind them is that predicting the effect variable using the cause variable should be 'simpler' than the reverse different notions of 'simplicity' giving rise to different criteria. We demonstrate the accuracy of the criteria on synthetic data generated under a broad family of causal mechanisms and types of noise.