We use China Family Panel Studies data to construct a four-party model of local government, healthcare providers, a service-integration platform, and older households and examine why medical spending rises after hospitalization while changes in household debt are not statistically detectable in the next survey wave. The model combines a Stackelberg game, quantal response equilibrium, and Markov state transitions to analyze protection payments, cost formation, service coordination, and household financing. The empirical results show that medical consumption and household out-of-pocket payments increase during hospitalization waves, whereas the effects on debt entry and debt persistence are not statistically detectable, and the debt-persistence estimate is imprecise. Under literature-supported directions and normalized parameters, the scenarios associate stronger protection, lower provider coordination costs, and higher net returns to platform accountability with lower household financing. Financing costs first affect financing strategies and then affect two-year debt states. By incorporating debt transitions into a multi-party framework for medical financial protection, the analysis connects protection payments, household financing, and financial states to clarify the observed debt-transition pattern.
This study examines how the structural characteristics of networks influence supply chain resilience under risk propagation. The study developed a probabilistic framework that integrates Bayesian networks with agent-based modeling. In this framework, recovery is modeled as a second propagation process that mirrors the disruption. Disruption cascades downstream, while restoration support flows upstream from customers. Three phase-specific metrics—probabilistic network robustness at onset, supply network structural recoverability at the peak, and a recoverability measure over the recovery window—are evaluated across 16,200 simulations spanning random, scale-free, small-world, and hybrid networks. Although the four topologies are statistically indistinguishable across all three metrics, regression analysis reveals systematic associations between structural characteristics and resilience that topology-level comparison alone obscures. These associations shift across resilience phases. Betweenness centrality is positively associated with all three metrics. Hub dominance shows its strongest positive association with initial robustness, and the alternative path index is negatively associated with PNR but positively associated with SNSR and RM. Disabling customer support halves realized recovery, confirming the substantive contribution of recovery propagation to overall resilience. Building on these findings, this study proposes an integrated classification framework comprising strategic and operational matrices that translates structural characteristics into actionable resilience management strategies.
Mid-market firms now share operational judgment with machine-learning systems while remaining too small to staff enterprise-grade governance. Viability theory and recent taxonomies of AI-era organizational pathologies explain why that combination is structurally demanding; they do not yet supply a diagnostic of fit between the human activity system and the model. This article describes the diagnostic. Following a design-science process on the practice-to-theory path, the Viable-System Alignment Model (VSAM) was specified by reanalysis of a theoretically sampled mid-market archive documented between 2003 and 2025. The original archive showed recurring and subsequently stable patterns concerning observation, authority, written judgment, and coordination load, which were reanalyzed as candidate conditions of human–AI fit. The artifact comprises Observability, Decision Rights, Codification, and Metabolic Load, each with a home in the Viable System Model and a socio-technical failure mode. The central proposition is that human–AI fit is a property of the viable system rather than of the software itself. Evaluation at this stage is analytical and descriptive rather than observational; independent field evaluation is reserved for subsequent research.
Social bots threaten the integrity of online social platforms by spreading misinformation, manipulating public opinion, and carrying out fraudulent activities. Detecting social bots is therefore not just a classification problem: looking ahead, a detector can also act as a discriminative component in agent-based social network generation, supplying realism-assessment signals for a “Generate–Detect–Explain–Refine” loop. Existing graph-based detection methods rely mainly on social relation graphs and do not explicitly treat behavioral similarity as a separate parallel graph branch. This paper proposes a dual-branch graph neural network for Twitter bot detection in which a heterogeneous social relation graph is accompanied by a sparse, conditionally active behavior-similarity auxiliary view; the paper’s main contributions are a competitive detector and a diagnostic protocol for attributing and repairing the value of that auxiliary view. The relation branch uses a Relational Graph Convolution Network to process a heterogeneous social relation graph, while the auxiliary branch applies a Graph Attention Network to a behavior similarity graph that links users with similar follower–following ratios, tweet type distributions, and post influence metrics. Experiments on TwiBot-22 show that the proposed model reaches 89.48 ± 0.07% accuracy and 78.11 ± 0.14% macro-F1 over five random seeds at a per-seed validation-calibrated decision threshold (89.92 ± 0.07% accuracy and 77.28 ± 0.31% macro-F1 at the default 0.5 threshold). Per-class and threshold analyses further show that the default operating point is precision-oriented on the minority bot class (bot recall 54.83 ± 1.60%), while operating near the validation-calibrated threshold (τ = 0.40) raises bot recall to 64.12 ± 1.31%. Ablation studies show that the semantic text embedding provides the largest single performance contribution and that the relation branch supplies the bulk of the graph signal. A degree-stratified analysis further shows that the behavior branch is functionally active on its ≈1% covered users—2.9× more prediction changes and 4.0× larger probability shifts than on uncovered nodes—where a validation-only subgroup threshold lifts macro-F1 from 51.85 to 67.17 without retraining, so that it can be turned into a deployable calibrated component.
