counterfactual explanations

Current what-if analysis lacks a unified conceptual framework, leading to terminological inconsistency across domains, structural ambiguity, and divergent interpretations. To address this, we conduct a systematic review of 141 papers in visual analytics and human-computer interaction, proposing Praxa—the first integrative framework that unifies scenario modeling, sensitivity analysis, and counterfactual analysis under a coherent paradigm. Praxa formally defines the underlying motivations, core components (hypothesis generation, intervention modeling, outcome evaluation), and a taxonomy of analytical types. It establishes a standardized terminology and structured model, exposing critical challenges including interpretability, causal modeling fidelity, and alignment with user intent. By clarifying conceptual boundaries and operational relationships among methods, Praxa significantly enhances cross-domain conceptual consistency and application clarity. The framework provides a rigorous foundation for theoretical advancement and the design of next-generation interactive analytical tools.

This study investigates the root causes of undesirable attributes—such as toxicity, negative sentiment, and political bias—in generative AI outputs, with a focus on the role of input prompts. To this end, it introduces the first counterfactual explanation framework tailored for non-deterministic generative models, termed Prompt Counterfactual Explanations (PCEs). By integrating a downstream classifier-guided counterfactual generation algorithm with targeted prompt perturbation strategies, the framework enables prompt-centric interpretability analysis. This approach overcomes key limitations of traditional explainable AI (XAI) methods when applied to generative models. Empirical results across three tasks—political stance, toxicity, and sentiment—demonstrate the framework’s effectiveness in generating meaningful PCEs, substantially enhancing prompt engineering efficiency and red-teaming capabilities, thereby facilitating precise identification and mitigation of harmful model outputs.

This study addresses a critical limitation in traditional reproducible research, where sharing only code and results fails to expose the implicit assumptions, expectations, and premises underlying an analyst’s reasoning—thereby hindering thorough evaluation of analytical quality. To overcome this, the paper proposes a formal modeling framework that explicitly translates the analyst’s tacit reasoning process into structured logical representations, statically capturing the construction logic of the analysis. This approach enables systematic scrutiny of the analytical chain of reasoning, assumption sensitivity, and conclusion robustness—even in the absence of the original data. Empirical validation on representative data analysis tasks demonstrates the framework’s effectiveness, achieving both logical visualization and data-free static assessment of analytical integrity.

Contemporary machine learning models’ complexity poses significant trust, regulatory, and ethical risks, yet existing explainability guidelines lack operational specificity. To address this gap, we conducted a controlled experiment with 124 developers, integrating cognitive process theory and sociological imagination to investigate how policy frameworks influence the design of end-user–oriented explanations for diabetic retinopathy screening models. Results reveal that all participants struggled to generate high-quality, policy-compliant, and empirically verifiable explanations; over 70% failed to accurately anticipate users’ comprehension barriers; and widely adopted technical methods (e.g., SHAP, Anchors) exhibit fundamental misalignment with real-world stakeholder needs. Our core contribution is identifying *developers’ inability to empathize with non-technical stakeholders* as the central mechanism underlying explanation failure—and proposing, for the first time, an empathy-centered educational intervention framework to bridge this gap.

Business professionals—non-technical domain experts—lack appropriate tools and methodologies for effective what-if analysis (WIA), hindering data-informed decision-making. Method: We conducted a two-phase mixed-methods user study—comprising contextual interviews and in-situ task-based evaluations—to systematically characterize their analytical behaviors for the first time. Contribution/Results: Based on empirical findings, we propose three domain-grounded design principles: business-contextual data preparation, risk-aware assessment, and domain-knowledge integration. We implemented and validated these principles in an interactive visual analytics prototype. The study identifies three critical support gaps, empirically confirms that six classes of what-if techniques significantly improve decision efficiency and confidence, and yields eight actionable design guidelines for commercial business intelligence systems. This work bridges a key theoretical and practical gap in WIA research concerning non-technical users.

This study addresses the lack of a systematic framework for identifying critical input variables and conducting sensitivity analysis under uncertainty in complex simulations, particularly in military decision-making contexts. The authors propose a unified sensitivity analysis framework that integrates local and global methods—including variance-based, derivative-based, screening, and uncertainty quantification techniques—and strategically maps these approaches to specific decision objectives such as factor prioritization, fixing, variance reduction, and mapping. Innovatively, the framework introduces a “sensitivity audit” mechanism to enhance traceability of model assumptions and promote responsible model usage. By providing a structured guide for high-dimensional, complex simulation systems, this work significantly improves model interpretability, transparency, and the credibility of decisions derived from such models.

Existing PresPM methods struggle to model temporal dependencies and causal effects of multi-stage interventions, either restricting decisions to single-step actions or relying on simulation/data augmentation—introducing a reality gap. This paper proposes the first backward-induction-based causal effect propagation framework that directly learns KPI-driven sequential intervention policies from observational event logs, eliminating the need for environment simulation. Our method integrates doubly robust estimation, propensity score weighting, and dynamic-programming-style backward induction to explicitly propagate and jointly optimize causal effects across interventions. Evaluated on synthetic and novel semi-synthetic real-world benchmarks, it significantly outperforms state-of-the-art methods, achieving 12.7%–23.4% KPI improvement. We further release a reproducible evaluation benchmark.

This study addresses the problem of clinicians’ overreliance on AI-driven decision support systems (DSS) and their diminished cognitive engagement by introducing two friction mechanisms—data-driven questioning and “what-if” counterfactual analysis—and empirically evaluating their impact on clinician reflection and decision-making in authentic clinical tasks. Leveraging a prototype DSS replicating real-world clinical workflows, the research integrates in-situ interviews and qualitative feedback from seven domain experts, analyzed through human-computer interaction and cognitive science frameworks. This work presents the first comparative assessment of these friction mechanisms in a real-world setting, demonstrating that “what-if” analysis enhances care quality, while data-driven questioning effectively directs clinicians’ attention to critical information. The findings further propose a novel perspective: leveraging AI-induced friction as a pedagogical tool for training novice clinicians.

This study addresses the challenge of intuitively representing complex interdependencies among simulation parameters in traditional tabular interfaces, which often lead to configuration errors and excessive cognitive load. To mitigate this, the work proposes the first application of an interactive Sankey diagram for visualizing parameter dependencies. A functional interface prototype was developed and evaluated against a conventional table-based approach using the PURE heuristic evaluation method, with a focus on user comprehension efficiency. Empirical results demonstrate that the Sankey diagram significantly reduces cognitive load by 51% and decreases interaction steps by 56%, thereby substantially enhancing the understandability and usability of configuration-intensive systems. This approach establishes a novel paradigm for visualizing parameter dependencies in complex simulation environments.