The research team, led by Nu Hoang and Thin Nguyen of Deakin University’s Applied Artificial Intelligence Institute, together with collaborators from the university’s Global Centre for Preventive Health and Nutrition, set out to solve a problem that has long undermined evaluations of community interventions. Traditional statistical approaches, such as linear regression, demand careful selection of covariates to ensure valid causal inference. Choosing the wrong variables, either omitting a genuine confounder or adjusting for an inappropriate one, can bias estimates of a program’s effect. Even with the right covariates, misspecifying the functional form of the model can still produce misleading conclusions. The team’s answer was not a new algorithm but a principled, end-to-end workflow that chains together causal discovery, formal identification through do-calculus, double machine learning, and interpretable subgroup discovery.

The first stage of the pipeline is causal discovery, the task of inferring cause-and-effect relationships directly from data rather than assuming them from prior theory. The researchers employed a hybrid causal discovery model designed for mixed-type data, meaning it can handle both continuous measurements, such as body mass index, and categorical variables, such as sex. The model works in two phases: a randomized conditional independence test builds the skeleton of the causal graph, and a cross-validation-based scoring function then orients and refines the edges. Because community intervention data are inherently longitudinal, the team adapted the model with temporal constraints, learning a graph over baseline variables first and then forcing edges to flow forward in time. This two-step design proved its worth in validation: it achieved significantly higher log-likelihood on held-out data than a one-step alternative, with a p-value below 0.0001.

The resulting causal graph is a map of how children’s characteristics and behaviors interconnect. Baseline age stood out as the most influential upstream variable, with the highest mean out-degree of 13.4 across bootstrap resamples, linking to physical activity, sedentary behavior, active transport, fruit intake, sweet beverage consumption, takeaway food consumption, and the outcome itself: the change in body mass index z-score over time. Baseline zBMI and sex were the next most recurrent upstream variables. Critically, the confounding pathways from sex, baseline age, and baseline zBMI to program participation appeared with a bootstrap frequency of 1.0, meaning they were perfectly stable across repeated resampling of the data. This stability gave the researchers confidence that the adjustment set identified from the graph could reliably strip away spurious correlations.

With the causal structure established, the team turned to double machine learning, a technique from econometrics that allows flexible machine learning models to be used in causal estimation without sacrificing statistical consistency. The method models both the outcome and the treatment assignment as functions of the covariates, then exploits two tricks to avoid the biases that normally plague machine learning estimators. Cross-fitting splits the data into partitions so that models are never evaluated on the data used to train them, guarding against overfitting bias. Orthogonalization, rooted in the classical Frisch-Waugh-Lovell theorem, regresses out nuisance functions and estimates the causal parameter from residuals, delivering a root-N consistent estimator free of regularization bias. Here, LightGBM models served as the nuisance learners, with hyperparameters tuned by fivefold cross-validated grid search.

The headline result of the estimation stage is a picture of profound heterogeneity. The average treatment effect across all children was a modest reduction of 0.02 in zBMI, with a 95 percent confidence interval spanning from minus 0.09 to plus 0.05, a range that crosses zero. But the individualized estimates told a far richer story, ranging from minus 0.25 to plus 0.15. Roughly 18.7 percent of children had confidence intervals entirely below zero, indicating a genuine benefit, while 8.7 percent had intervals entirely above zero, suggesting the program may have been counterproductive for them. The remaining 72.6 percent had intervals that included zero, reflecting the substantial uncertainty inherent in individual-level causal estimates. The researchers are careful to stress that these individualized figures describe broad patterns of heterogeneity rather than decision-grade predictions for any single child.

To probe what drives this variation, the team first examined linear correlations between the estimated effects and baseline variables. Age showed the strongest association, with a Pearson coefficient of 0.59, while no other baseline variable correlated significantly. Because linear correlation can miss nonlinear structure, the researchers then trained a shallow decision tree to classify children into positive-effect, negative-effect, and no-effect groups based on nine baseline characteristics. The tree partitioned the data using just two variables: age and weekly takeaway food consumption. Children under 8.9 years old formed the most robustly benefited subgroup, with a dominant-class probability of 0.93 and a mean individual treatment effect of minus 0.096, more than four times the pooled average. Children aged 8.9 to 13.6 who ate takeaway food less than once per week also benefited, with a mean effect of minus 0.038. In contrast, adolescents above 15.2 years showed a clearly negative response, with a mean effect of plus 0.018.

The robustness of these findings was tested from multiple angles. Bootstrap validation of the decision tree across 100 resampled datasets confirmed that age was selected as a splitting variable in every single run, while takeaway food consumption appeared in 40 percent of runs and no other variable appeared at all. Refutation tests bolstered the causal estimate itself: a placebo treatment test, which replaces the real treatment with random values, yielded an effect of minus 0.0038 that was statistically indistinguishable from zero, exactly as expected if the original estimate reflects a genuine causal relationship. Subset validation produced an effect of minus 0.016, not significantly different from the original. An E-value sensitivity analysis indicated that an unmeasured confounder would need to be associated with both treatment and outcome by a risk ratio of at least 1.23 to explain away the average effect, though the subgroup effect for younger children is substantially larger and more resilient.

The authors are candid about the limitations of their work. Communities self-selected into the intervention programs, so unmeasured factors such as local political support or socioeconomic resources could confound the results, potentially overstating benefits. Spillover effects, in which children in comparison communities are indirectly exposed to intervention activities, could bias estimates toward the null. The pooled average effect is small and its confidence interval crosses zero, so the researchers emphasize that the study’s main contribution lies in demonstrating how heterogeneity can be identified and characterized, not in claiming a large overall benefit. The individualized estimates are explicitly framed as exploratory and potentially model-dependent, useful for revealing broad subgroup structure rather than for clinical decision-making at the level of a single child.

Even with those caveats, the implications are considerable. If community-based obesity prevention is genuinely most effective for younger children and least effective for older adolescents, then policymakers have a concrete, actionable signal: interventions may need age-sensitive redesign, with different strategies for teenagers than for primary-school children. The interaction analysis adds nuance, finding that the combination of age and takeaway food consumption drives heterogeneity beyond age alone, and that physically active children who eat fewer takeaway meals benefit more. Beyond obesity, the framework is explicitly generalizable. The authors argue that the same workflow, causal discovery to build the graph, do-calculus to identify the adjustment set, double machine learning to estimate effects, and tree-based subgroup analysis to interpret them, could be applied to any complex community intervention where average effects conceal the individuals the program actually helps. In an era when public health budgets are finite and one-size-fits-all programs increasingly look inadequate, that may be the study’s most viral-worthy message: the average is a lie, and the tools to see past it now exist.

Subject of Research: Causal machine learning analysis of heterogeneous treatment effects in community-based childhood obesity prevention interventions

Article Title: Causal machine learning for understanding heterogeneous effects of childhood obesity prevention

Article References: Hoang, N., Nguyen, T., Duong, B., Nichols, M., Brown, V., Backholer, K., Allender, S., & Nguyen, T. (2026). Causal machine learning for understanding heterogeneous effects of childhood obesity prevention. International Journal of Data Science and Analytics, 22(1), Article 326. https://doi.org/10.1007/s41060-026-01156-z

Image Credits: AI Generated

DOI: 10.1007/s41060-026-01156-z

Keywords: causal machine learning, causal discovery, causal inference, childhood obesity, community-based interventions, heterogeneous treatment effects, double machine learning, BMI z-score, public health, age heterogeneity, subgroup analysis, preventive health

The new method, called CausalCACTUS, is described in an open-access paper published in the journal Machine Learning on 28 September 2026. It extends an earlier framework known as CACTUS, which generated counterfactual explanations in a compressed latent space learned by a variational autoencoder, while allowing users to specify contextual constraints such as age or family status that the explanation must respect. CausalCACTUS adds what CACTUS lacked: an explicit model of cause and effect among the input features, learned automatically from data rather than supplied by human experts.

The technical core of the approach rests on three pillars. First, the researchers use a causal discovery algorithm called ReX, which trains predictive models for each feature and applies bootstrapped Shapley value attributions to identify which features depend on which. The output is a directed acyclic graph, a map of parent-child relationships such as the chain linking revolving credit utilization to the number of credit lines and then to a person’s debt ratio. Second, this graph is converted into dynamic feasibility constraints during the counterfactual search: a child feature is only allowed to change if at least one of its parents changes as well. Third, a phase-based optimization strategy first drives the search toward any valid counterfactual and then refines it for contextual alignment and minimal distance, with a weighting parameter that gradually shifts emphasis from prediction validity to context preservation.

The counterfactual search itself takes place not in the original feature space but in a four-dimensional latent space learned by a beta-variational autoencoder. This composite model includes an auxiliary classification head that predicts the user-defined context from the latent representation, allowing the optimizer to penalize candidates that drift outside the specified context, for example an applicant older than fifty. A causal distance loss, computed with an adaptive binary mask that is updated at every iteration, penalizes changes to features whose causal parents have not yet been modified. Feature values are also clipped to ranges observed in the training data after each decoding step, keeping every candidate within plausible bounds.

The evaluation covered five publicly available tabular datasets spanning credit risk, income prediction and legal education, ranging from 462 to 16,668 instances, with binary context variables derived from demographic and financial attributes. Against four baseline methods, including LatentCF++, two variants of PrototypeCF and the original CACTUS, CausalCACTUS ranked among the top two on validity in eight of ten comparisons and improved validity over the original CACTUS from roughly 0.7 to around 0.9. More strikingly, it achieved the best context-alignment scores, measured by a local outlier factor computed within the target context, in nine of ten scenarios, and the best causal-compact consistency, a metric combining causal constraint satisfaction with sparsity of changes, in nine of ten configurations.

Statistical analysis reinforced these findings. A Friedman test followed by a Wilcoxon-Holm post-hoc analysis across ten benchmark tasks showed that CausalCACTUS achieved the best average ranks, between 1.0 and 1.4, on proximity, context alignment, compactness and causal consistency. Ablation studies revealed a division of labor between the two new components: the phase-based optimization primarily boosted validity and compactness, while the causal feasibility constraints mainly improved context alignment and causal consistency. The researchers also mapped how the loss-weighting parameter alpha trades validity against context alignment, identifying a balanced setting near 0.7.

A concrete example from the credit dataset illustrates why the framework matters. For a loan applicant over fifty initially classified as high risk, competing methods proposed income changes ranging from a decrease of more than 15,000 dollars to increases exceeding 5,000 dollars, figures the authors describe as implausible for someone at that life stage. CausalCACTUS instead suggested a moderate adjustment of 1,605 dollars, required fewer overall feature modifications, and respected the learned dependency chain between credit utilization, credit lines and debt ratio, which the baseline methods violated. Correlation analysis of feature changes across test instances showed that CausalCACTUS produced strong correlations concentrated on causally related feature pairs, closely matching the learned graph, while baselines either showed weak correlations or correlations misaligned with causal structure.

The authors are candid about the limitations of their approach. Because it relies on a causal graph estimated by a discovery algorithm rather than a fully specified structural causal model, missing edges in the estimated graph could introduce inaccurate dependencies into the search; expert-defined graphs can substitute where reliable domain knowledge exists. The generative model is also a bottleneck: reconstruction errors or misalignment between the latent representation and the underlying causal mechanisms may cause information loss or unintended feature changes. Future work, the team suggests, could replace the beta-variational autoencoder with diffusion models, normalizing flows or causally informed generative variants, and could jointly learn latent representations and causal structures with uncertainty estimates attached to the learned graph.

The implications reach well beyond credit scoring. Counterfactual explanations have been linked to the right to explanation under the General Data Protection Regulation, and as algorithmic decisions spread through finance, hiring and education, the gap between mathematically valid explanations and practically actionable ones becomes a matter of public trust. By showing that causal consistency and user-defined context can be enforced simultaneously in a latent-space search, without demanding a fully specified causal model as a prerequisite, CausalCACTUS offers a template for explanations that a rejected applicant, a loan officer and a regulator could all accept. The source code has been released publicly on GitHub, and the authors point toward extensions to time-series classification and sequential prediction as the next frontier for causally grounded, context-aware explainability.

Subject of Research: Causally consistent and context-aware counterfactual explanations for machine learning classifiers

Article Title: CausalCACTUS: A Causally Consistent and Context-Aware Framework for Counterfactual Explanations

Article References: Wang, Z., García, D., Enguita, J. M., Chatterjee, S., & Jansson, M. (2026). CausalCACTUS: A Causally Consistent and Context-Aware Framework for Counterfactual Explanations. Machine Learning, 115(10), Article 233. https://doi.org/10.1007/s10994-026-07155-2

Image Credits: AI Generated

DOI: 10.1007/s10994-026-07155-2

Keywords: counterfactual explanations, explainable AI, causal discovery, machine learning, variational autoencoder, structural causal models, latent space optimization, algorithmic recourse, model interpretability, context-aware learning, tabular data, GDPR

The authors argue that the field’s fixation on efficacy has deep historical roots. Behavioral science grew up alongside public policy, where cost-effectiveness was the primary concern, and that legacy shaped the questions researchers asked. Large meta-analyses have found that individually focused behavioral interventions often show limited effectiveness and weak persistence, prompting some scholars to call for system-level approaches instead. Veltri and Banerjee contend that this critique stems in part from three unresolved problems: the inability of most studies to reveal mechanistic explanations of intervention delivery, insufficient attention to heterogeneity in intervention effects, and neglect of temporal dynamics. Their paper targets the first of these, because mechanistic understanding is what allows theories to advance, interventions to be fine-tuned, and results to transfer across contexts.

The conceptual core of BriDGE rests on two statistical pillars. The first is the directed acyclic graph, or DAG, a graphical formalism in which nodes represent variables — cognitive processes, emotional states, social influences — and directed edges represent hypothesized causal influences. The acyclic structure enforces unidirectional, time-ordered relationships and rules out feedback loops that would complicate inference. Where conventional moderation and mediation models test a small set of focal pathways, DAGs encode the broader causal structure: they expose potential confounders and colliders and show which variables must, or must not, be adjusted for to identify a causal effect. By making assumptions explicit before any data are collected, DAGs serve as a blueprint for both study design and analysis, formalizing the entities and activities that together constitute a behavioral mechanism.

The second pillar is the generalized additive model, or GAM. Behavioral mechanisms rarely behave like straight lines. A mediator may exert a threshold effect on an outcome, or its influence may increase at a diminishing rate, and linear regression compresses all of that curvature into a single average slope. GAMs solve this by fitting smooth, data-adaptive functions — thin-plate regression splines in the implementation used here — that capture nonlinear main effects, mediator–mediator interactions, and treatment-by-mediator effect modification without imposing rigid parametric assumptions. The authors implement GAMs through the mgcv software framework, with smoothing parameters estimated by restricted maximum likelihood and basis dimensions checked through standard diagnostics. The combination is deliberate: DAGs specify which causal routes might exist, while GAMs quantify the functional shape of each route.

BriDGE unfolds in six steps. First, the researcher constructs a DAG representing hypothesized relationships among intervention, mediators, and outcomes. Second, data are collected with mediators and outcomes measured at appropriate times and with sufficient precision — a requirement that conventional trials often fail to meet, since mechanistic analyses demand larger samples and more granular measurement than efficacy testing alone. Third, causal discovery algorithms are applied to interrogate the hypothesized graph empirically. The protocol uses the Max-Min Hill-Climbing algorithm, constraining the search so that edges from the randomized treatment to mediators are whitelisted and any edges pointing back into treatment are blacklisted, thereby preserving the logic of random assignment. Crucially, the discovered graph is accompanied by bootstrap arc-strength summaries, and unstable edges are treated as ambiguous hypotheses rather than definitive causal claims.

The fourth step fits GAMs to capture nonlinear and interactive relationships, and the fifth conducts mediation analysis to decompose the total treatment effect into direct and indirect components. Because multi-mediator settings with interactions make simple path-by-path attribution fragile, the protocol emphasizes interventional estimands: the interventional direct effect, which captures what remains when the mediator distribution is held at its control regime, and the joint interventional indirect effect, which summarizes how much the outcome changes when the entire joint mediator distribution shifts from control to treated conditions. The sixth step quantifies uncertainty through nonparametric bootstrapping with bias-corrected and accelerated confidence intervals — at least 1,000 replicates, with 2,000 used in the paper’s simulations — alongside sensitivity analyses probing discretization choices, model specification, mediator measurement error, and potential mediator–outcome confounding.

To validate the pipeline, the authors ran a fully synthetic simulation with 1,000 units in which the true causal graph and all target effects were known by design. The simulated outcome included a quadratic mediator effect, mediator–mediator interaction, and treatment-by-mediator effect modification — precisely the non-additive structure that linear mediation models cannot represent. In a comparison against conventional linear mediation workflows, the traditional approach inflated mediator-specific indirect effect estimates because curvature and synergy were compressed into average slopes and then misallocated across pathways. The practical consequence, the authors note, is that applied researchers can end up concluding that a single psychological construct “explains” an effect when the true mechanism depends on thresholding, saturation, or mediator synergy — and then refine interventions around components that are not causally decisive.

The team also built a semi-synthetic benchmark on the scaffold of the JOBS II randomized field experiment, retaining its real randomized treatment assignment and baseline covariates while generating mediators and outcomes from a known structural model. This preserves the skewness, heaping, and baseline correlations typical of behavioral data while keeping the ground truth intact. The results were striking: an interaction-capable outcome model closely recovered the true interventional direct effect, joint indirect effect, and total effect, with confidence intervals covering the known values. An additive-only specification, by contrast, inflated the decomposition components, and its direct-effect interval excluded the true value — even though the estimated total effect looked similar. In other words, the headline result of a trial can appear robust while the mechanistic story beneath it silently changes depending on whether key interactions were modeled.

The authors are candid about limitations. Randomization guarantees that treatment is exogenous, but mediation identification additionally requires no unmeasured confounding of mediator–outcome relationships — an assumption that cannot be tested and that is often questionable in behavioral research, where demand characteristics, common-method bias, or latent motivation may jointly influence mediator reports and behavior. Causal discovery from finite samples is inherently uncertain, with multiple graphs often observationally equivalent, so BriDGE treats discovered structures as hypothesis generators rather than truths. Data requirements are substantial: high-quality mediator measurement, low attrition, and samples large enough for stable discovery and precise inference. The protocol therefore includes a decision flowchart for when to use BriDGE versus simpler single-mediator or linear models, and simulation-based power planning tools, since indirect effects attenuate quadratically as effect sizes shrink.

Everything is packaged for adoption. The workflow is implemented in an open-source R package, bridgeR, covering DAG specification and validation, constrained causal discovery, GAM fitting, and bootstrap-based mediation, with benchmarking templates for runtime and power available through an Open Science Framework repository. The authors frame the protocol as a complement to recent calls for integrative experimental designs that systematically sample from a broader design space, and they argue that policymakers stand to gain the most: knowing not just that an intervention works, but how and why, allows strategies to be adapted to local contexts, resources to be allocated efficiently, and behavioral theories to become genuinely cumulative. The next critical step, they acknowledge, is applying BriDGE to fully empirical trial data, where the true causal graph is unknown and missingness, measurement error, and construct-validity constraints are the norm rather than the exception.

Subject of Research: A methodological protocol integrating causal graphs and generalized additive models to reveal mechanisms in behavioral intervention experiments

Article Title: BriDGE the gap: Improving behavioral research by integrating DAGs and GAMs into experiments

Article References: BriDGE the gap: Improving behavioral research by integrating DAGs and GAMs into experiments. (n.d.). https://doi.org/10.3758/s13428-026-03146-2

Image Credits: AI Generated

DOI: 10.3758/s13428-026-03146-2

Keywords: behavioral science, causal inference, directed acyclic graphs, generalized additive models, randomized controlled trials, mediation analysis, causal discovery, research methods, behavioral public policy, open-source software, simulation study, intervention design