Generative artificial intelligence (AI) complicates academic integrity by blurring the boundary between assistance and authorship, enabling cognitive delegation, and introducing algorithmic mediation into assessment and institutional decisions. This article presents a critical conceptual synthesis, not a systematic review. A purposive corpus of 64 academic and policy sources was examined through comparative coding, negative-case analysis, source-to-concept tracing, and normative interpretation. The synthesis distinguishes integrity of academic work, integrity of assessment, and institutional procedural integrity. It proposes an integrated governance framework connecting preventive governance, ethical AI literacy, governed verification, restorative accountability, and proportionate discipline. The framework argues that policy clarity must reach the assessed task; AI literacy supports judgement but cannot neutralize strategic misconduct; automated indicators require corroboration, competent human review, reasons, and appeal; and restorative processes are appropriate only when harm, affected parties, voluntary participation, responsibility, repair, and reintegration are substantively addressed. Its principal contribution is a case-to-governance feedback mechanism through which integrity cases generate institutional learning and trigger revision of policy, assessment design, literacy provision, and technology oversight. Seven testable conceptual propositions, a response-selection pathway, role-specific duties, staged implementation responsibilities, and evaluation indicators are provided. The framework remains a normative and empirically testable proposal rather than a validated intervention or universally effective solution.

Height systems form a mathematical interface between physical geodesy, engineering surveying, and digital cartography. Although satellite positioning efficiently provides ellipsoidal heights, practical infrastructure, mapping, hydrological, and monitoring tasks require gravity-related heights that are compatible with national vertical datums. This paper develops a denominator-based framework in which dynamic, orthometric, and normal heights are interpreted as metric realizations of a common geopotential number. Starting from the line-integral definition of geopotential, the principal height formulae are derived; first-order sensitivities to geoid undulation, height anomaly, and mean gravity are established; and uncertainty propagation is analyzed. A Romania-oriented computational experiment, explicitly defined as a representative model-behavior study rather than an official national adjustment, uses lowland, plateau, and mountain-influenced settings consistent with the Constanta and Black Sea 1975 normal-height context. The results show that modeled normal-orthometric separations remain below 3 mm in representative low-relief locations but increase to 17.1 mm in Suceava, 29.5 mm in Cluj-Napoca, and 85.6 mm in the mountain-influenced Brasov case. The dynamic-normal differences remain small at low elevations but become systematic where the normal-gravity denominator departs from the selected reference value. The Monte Carlo experiment indicates standard uncertainties of approximately 4.2–4.4 cm for normal heights when a 1.5 cm ellipsoidal-height uncertainty and a 4.0 cm quasi-geoid uncertainty are assumed. A single-point covariance example gives 31.7 mm for normal-height conversion and 36.9 mm for orthometric-height conversion under the stated correlation assumptions. The transformation-surface comparison further shows that a quadratic local model reduces leave-one-out cross-validation error from 16.73 mm to 9.95 mm relative to a planar model in the synthetic Romania-oriented scenario. The study concludes that the height-system label must be treated as part of the mathematical model and metadata, and it proposes a geopotential-centered computational pathway for survey adjustment, uncertainty control, and metadata-safe geospatial export.

Natural hazards can disrupt transport infrastructure through spatially heterogeneous flooding, seismic shaking, terrain conditions, and observed ground motion, while local exposure alone does not determine system-wide consequences. This study develops a reproducible multi-hazard Graph Artificial Intelligence framework for infrastructure resilience screening on the Romanian strategic road network. A frozen attributed line graph combines network topology with European land-cover, terrain, seismic, flood, and ground-motion evidence, and an independent disruption simulator provides continuous resilience-loss targets. GCN, GraphSAGE, and GAT were trained under a leakage-controlled three-seed protocol. On the blind 194-scenario PRIMARY TEST, ensemble RMSE/R² were 0.013343/0.997104 for GCN, 0.013978/0.996822 for GraphSAGE, and 0.015830/0.995924 for GAT. A fresh GCN branch achieved RMSE/R² = 0.018230/0.995852 on 111 component-disjoint cases and 0.010468/0.991654 on 256 severity-specific held-out cases. On 208 predefined external anchors held out from Graph-AI fitting and selection, GAT attained the lowest ensemble RMSE (0.021620; R² = 0.975065), while seismic-only transfer remained weak (n = 10; R² = 0.251–0.386). A frozen six-family GCN ablation (18 checkpoints; 194 matched test scenarios) found RMSE = 0.026175 after seismic feature omission versus 0.013343 for the full-input GCN; this is simulator-specific sensitivity, not causal hazard attribution. The results support accurate emulation of the frozen simulator and transfer-aware screening, not structural safety certification. The model outputs are simulator-based rather than observed disruption outcomes; neither an end-to-end simulator-to-GNN speedup nor calibrated prediction intervals are established by the present evaluation.