attribution modeling

This study addresses the lack of formal definitions in existing root cause analysis methods, which are often limited to root nodes in causal graphs or biased toward proximate causes. Within the potential outcomes framework, this work proposes the first counterfactual definition of root cause at the individual level and introduces a probabilistic measure—Probability of Root Condition (PRC)—to quantify the likelihood that a candidate set of variables constitutes a root cause for a specific outcome. Under standard causal assumptions, the authors derive an explicit identification formula for PRC by integrating causal mediation analysis with counterfactual reasoning, thereby establishing its identifiability. The effectiveness and practical utility of the proposed approach are demonstrated through two numerical examples, filling a critical gap in the formal theory of root cause analysis.

Traditional accountability frameworks fail in AI system accidents due to high technical interconnectivity, ethical ambiguity, inherent uncertainty, and regulatory gaps. Method: This paper proposes Computational Reflective Equilibrium (CRE), the first formalization of the philosophical reflective equilibrium model as a computable responsibility attribution mechanism; it innovatively introduces an assertion activation control mechanism to enable sensitivity analysis and iterative refinement of responsibility distributions. Grounded in computational philosophical modeling, weighted graph-based equilibrium solving, and formal assertion logic, CRE supports continuous monitoring, reflective adjustment, and institutional optimization of responsibility allocation. Contribution/Results: Evaluated in an AI-assisted clinical diagnosis simulation, CRE demonstrates strong interpretability, ethical consistency, and dynamic adaptability—effectively reconciling technical complexity with normative accountability requirements in socio-technical AI systems.

This study addresses the problem of quantitatively attributing causal responsibility to two binary risk factors following an adverse outcome. The authors propose a framework for average causal responsibility based on the distribution of latent causal types. Under assumptions of no confounding and monotonicity, they achieve nonparametric identification of this responsibility through structural balancing conditions—a result established here for the first time. When these assumptions fail to hold, the method yields sharp bounds instead. Integrating potential outcomes, causal type analysis, and counterfactual reasoning, the approach is successfully applied to the canonical case of lung cancer jointly caused by smoking and asbestos exposure, enabling a quantitative decomposition of their respective causal contributions.

Real-world root cause analysis (RCA) faces a critical challenge: post-intervention distributions often contain only a few—or even a single—sample, rendering distribution-dependent or low-density-region regression methods statistically ill-posed. This paper proposes a lightweight root cause identification framework that requires neither counterfactual reasoning nor a fully specified structural causal model (SCM). It operates either given a causal DAG or, in the absence of one, solely from an anomaly score ranking. We theoretically prove that low-scoring anomalies rarely trigger high-scoring ones and derive a probabilistic upper bound on non-monotonic propagation paths. By abandoning Shapley-value-based attribution and density-sensitive regression, our method achieves linear time complexity O(n). It eliminates SCM fitting and counterfactual computation while providing rigorous theoretical guarantees and strong empirical performance.

This work addresses a critical limitation in existing root cause analysis methods for anomalies: their failure to distinguish between two fundamentally distinct sources—measurement errors and mechanism shifts—often leading to misdiagnosis. To resolve this, the paper proposes the first causal framework that explicitly models both anomaly types by treating them as implicit interventions on latent “true” variables and observed “measured” variables. A structural causal model (SCM) with latent variables is constructed, and maximum likelihood estimation is employed to simultaneously classify anomaly types and localize root causes. Theoretically, the approach is shown to be identifiable without requiring prior knowledge of the causal graph structure. Empirical evaluations demonstrate state-of-the-art performance in root cause localization, accurate anomaly-type classification, and robustness even when the underlying causal graph is unknown.

This study addresses the problem of interactive causal attribution in retrospective causal inference, where two binary exposures jointly influence a binary outcome. By introducing posterior probabilities to quantify the individual contributions of each exposure and their interaction, the work establishes, for the first time within a randomized controlled trial framework, identifiable conditions for interactive causal attribution—overcoming the limitations of traditional approaches that struggle with retrospective interaction effects. The proposed method integrates Bayesian posterior modeling, causal graphical models, and identifiability theory, leveraging auxiliary secondary outcomes observed after the primary outcome to achieve parameter identification. Applied to the classic case of lung cancer caused by smoking and asbestos exposure, the analysis reveals that the disease is primarily driven by the synergistic interaction between the two exposures rather than by either exposure alone.

Traditional human factors analysis relies on correlation-based testing, which only addresses descriptive “what is” questions and fails to resolve causal “if–then” queries. It remains vulnerable to confounding and collider variables, leading to biased decision-making. This paper proposes a paradigm shift from descriptive analysis to causal inference. Leveraging post-occupancy evaluation data from built environments, we integrate causal discovery algorithms, structural equation modeling, and counterfactual intervention analysis to construct directed causal networks among variables—explicitly distinguishing descriptive from interventional queries. Our key contribution is the first systematic application of a rigorous causal inference framework to the human factors domain, enabling robust identification of intervention priorities and hierarchical causal pathways. Empirical validation demonstrates that this approach significantly enhances the accuracy and scientific rigor of human factors–driven system optimization decisions, establishing a novel, interpretable, and actionable causal analysis paradigm for human factors engineering.

This work addresses the challenge that causal analysis methods, due to their conceptual complexity and limited validation on real-world data, remain difficult for domain experts to use effectively. To bridge this gap, we propose ORCA—the first end-to-end, interactive causal analysis collaborator designed specifically for non-expert users. ORCA employs a multi-agent architecture to jointly interpret user intent and supports the full causal analysis pipeline, including causal discovery, effect estimation, interpretability analysis, and root cause diagnosis, with adjustable levels of automation ranging from fully automatic to highly manual intervention. The system automatically generates structured reports, visualizations, and performance comparisons, substantially lowering the barrier to entry. Experimental evaluations across multiple real-world scenarios demonstrate that ORCA significantly enhances the efficiency, accuracy, and accessibility of causal analysis.