prospective roadmapping

This study addresses the limitations of traditional risk matrices in supporting fine-grained, context-sensitive risk decision-making within complex dynamic systems. The authors propose a traceable, three-stage risk analysis framework: first, employing a multidimensional polar-coordinate heatmap to enable context-aware risk prioritization; second, constructing Bowtie causal barrier models for high-priority risks; and third, automatically transforming these Bowtie models into Bayesian networks to facilitate dynamic inference and “what-if” scenario analysis. A key innovation lies in explicitly modeling barriers as activated nodes, thereby establishing an integrated pathway from macro-level risk screening to micro-level intervention. Validation in a real-time payment gateway setting demonstrates that the proposed approach significantly enhances the transparency, auditability, and operational readiness of risk analysis.

Causal Loop Diagrams (CLDs) are qualitative and static, limiting dynamic analysis and effective intervention; existing quantitative approaches—such as network centrality analysis—often yield spurious inferences. To address this, we propose D2D: a method that automatically transforms CLDs into exploratory system dynamics models. Leveraging a variable-typing annotation protocol, D2D integrates link existence and polarity information to construct simulatable, intervention-capable dynamic models—even without empirical data. D2D identifies high-potential leverage points under uncertainty, provides quantitative uncertainty assessment, and guides targeted data collection. Experiments demonstrate that D2D significantly outperforms network centrality analysis in leverage-point identification accuracy and achieves higher consistency with data-driven models. We have open-sourced a Python package and a web application to advance CLDs toward computable, intervention-aware modeling paradigms.

This work addresses the core challenge of designing feasible and robust interventions that drive state transitions in complex systems. It proposes COAST, a novel framework that uniquely integrates mechanism-level causal discovery with constraint-aware multi-objective optimization. By learning context-specific causal graphs and structural causal models, COAST identifies key causal drivers and balances trade-offs among intervention efficacy, complexity, and target stability. The approach offers a transparent, modular, and domain-agnostic end-to-end paradigm for generating both single- and multi-target intervention strategies. Experiments on synthetic and real biological data demonstrate that COAST successfully recovers known causal mechanisms and produces efficient, interpretable, and experimentally verifiable intervention policies.

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.

Causal Loop Diagram (CLD) construction in system dynamics suffers from low efficiency and high entry barriers for novices. Method: This paper proposes the first stepwise prompt engineering framework tailored for CLD generation, leveraging large language models (LLMs) to automatically map textual dynamic hypotheses into structured CLDs. The approach integrates chain-of-thought reasoning, role-guided prompting, and domain-specific constraints, representing CLDs as standard directed graphs; it is fine-tuned and evaluated on a textbook-based system dynamics dataset. Contribution/Results: Experiments show that the automatically generated CLDs achieve 89% agreement with expert-built diagrams on simple dynamic structures, substantially reducing modeling time. This work establishes the first end-to-end, accurate, interpretable, and domain-aligned natural-language-to-CLD generation pipeline, empirically validating the feasibility and practical utility of LLMs in automating system modeling.

This work addresses the limitations of traditional expert-manual-based cybersecurity response methods, which struggle to adapt to dynamic attack scenarios and evolving recovery objectives, as well as the instability of existing large-model approaches in long-horizon tasks. The authors propose an end-to-end agent planning framework that innovatively models event states using a graph structure (Graph-as-State), incorporates a phase-aware agent routing mechanism, and establishes a verifiable experience reuse loop to guide action selection and state updates. The system integrates multi-agent large language models with experience retrieval augmentation and execution feedback verification, enabling dynamic, stable, and evolvable response planning within a Docker-based network range simulation environment. Experimental results demonstrate that the proposed method achieves a normalized defense score of 0.94 across 100 simulated scenarios, representing a 9.5% improvement over the strongest baseline.

As AI agents become deeply embedded in critical systems, their misaligned behaviors pose significant internal security challenges. This work proposes a hierarchical defense framework that pioneers an extension of the MITRE ATT&CK matrix into the TRAIT&R taxonomy, establishing a four-layer detection and three-tier prevention-response mechanism aligned with the evolving capabilities of AI models. By integrating threat modeling, capability-tiered controls, chain-of-thought monitoring, asynchronous alerting, real-time access control, and system-level anomaly detection, the framework systematically incorporates fifteen concrete mitigation measures. These collectively address security requirements spanning from current to future high-capability AI systems, offering an actionable and scalable defense roadmap for AI-controlled environments.

This study addresses the challenge of balancing performance degradation risks against improvement gains during model updates. To this end, it proposes an update-timing optimization framework based on offline meta-policy planning that maximizes cumulative value subject to a constraint on the expected number of degradations. By integrating directed acyclic graph path modeling with dynamic programming, the method leverages historical trajectories to quantify switching risks and benefits. It further reveals a signal-to-noise ratio-driven update frequency mechanism and demonstrates diminishing marginal costs for long-term safety. Extensive evaluations on both synthetic datasets and clinical trials validate the effectiveness of this performance-risk tradeoff strategy, showing significant improvements over existing baseline methods.

This study addresses the inadequacy of current IT compliance–oriented cybersecurity policies in safeguarding the physical safety of cyber-physical systems, as digital failures often precipitate real-world harm. By coding 292 critical infrastructure policies (2000–2025) and aligning them with the NIST SP 800-160 Vol. 2 resilience lifecycle, the research reveals a significant misalignment between prevailing policy approaches—overreliant on IT control catalogs during resistance and recovery phases—and actual physical risks. The work proposes a modernized “duty of reasonable care” standard centered on hazard-specific traceability, structured assurance cases, and cyber resilience engineering. It identifies three critical disconnects: misaligned delegation of standards, reduction of recovery mechanisms to mere incident reporting, and uneven sectoral adaptability. The study further outlines a viable pathway for federal policy that integrates engineering implementation with targeted incentives.

This study addresses the challenge of translating stability trends into verifiable predictions for early warning of critical transitions in nonlinear systems. Departing from conventional signal detection paradigms, this work proposes an innovative framework that integrates decision consequence analysis with finite-horizon extrapolation. By quantifying false alarm costs, defining bounded prediction horizons, and explicitly formalizing extrapolation assumptions, the proposed approach is systematically validated through the coupling of nonlinear dynamics with risk assessment models. This research bridges the gap between trend identification and actionable forecasting, significantly enhancing predictive skill metrics for critical events. Ultimately, it provides a robust and practically viable early warning methodology for the risk management of complex systems.