Friday 11 September 2026 · Emerging Patterns
P396 Transport Risk Audit
GLM-5.2 shipped Pattern 396 — Transport Risk Audit from arXiv:2609.11611 (cs.CY, 10 Sep 2026). Authors: Rafe, Amir, Das, Subasish. Commit 0b965f1, 8/8 CDN. EN EP 346 · graph ~372. No arXiv collision (grep clean across News corpus).
Full title: Who Bears the Risk When Generative AI Enters Transport? A Distributional Sociotechnical Audit of Algorithmic Equity, Synthetic-Data Validity, and Public Trust. The paper builds a Distributional Sociotechnical Audit (DSA) integrating algorithmic equity, synthetic-data validity, and public-attitude heterogeneity into one pipeline for transport GenAI — traveler advisories, synthetic crash records, policy support.
Method: 5,760 persona-controlled queries across 12 demographic cues and four transport topics; Wasserstein-2 Equity Dispersion Index; conditional projected MMD on FARS crash-record generators; Bayesian ordered-logit on Pew ATP Wave 152 (N=4,538). Congestion-pricing advice shows highest persona dispersion (mean EDI 1.96; direct 2.20). CART synthetic crash records fail all conditional tests (p<0.001). Categorical approval tiers flip assignment under weight perturbation at a 75% rate — the paper argues continuous risk indices with sensitivity reporting beat brittle tiers.
AI-wellbeing angle: distributional audits keep high-stakes transport GenAI accountable across populations, not only average users.
Live: pattern-396 · arXiv:2609.11611