Collective Recourse for Generative Urban Visualizations

2025-09-14

AI Summary

Text-to-image diffusion models exacerbate societal biases in urban future visualization, undermining planning equity. To address this, we propose the first collective remediation framework embedded within the diffusion model stack, establishing an end-to-end “visual vulnerability reporting–triage–remediation–validation–closure” pipeline. We introduce a novel authorization scoring threshold—based on severity, saturation, representativeness, and evidentiary support—that achieves 93% precision and 75% recall at threshold 0.12. Our method comprises four remediation primitives (e.g., reverse/negative prompting, dataset editing, reward model tuning) and governance tools (e.g., interactive dashboard, rotating review panel). Empirical evaluation across 240 cases shows prompt-level fixes are fastest (median 2.1–3.4 days) but exhibit high recurrence (21%–38%); dataset and reward model interventions yield greater durability (recurrence 12%–18%) and higher planner adoption (30%–36%).

Technology Category

• Computer Vision: Diffusion Models for Vision
• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Text-to-image diffusion models help visualize urban futures but can amplify group-level harms. We propose collective recourse: structured community "visual bug reports" that trigger fixes to models and planning workflows. We (1) formalize collective recourse and a practical pipeline (report, triage, fix, verify, closure); (2) situate four recourse primitives within the diffusion stack: counter-prompts, negative prompts, dataset edits, and reward-model tweaks; (3) define mandate thresholds via a mandate score combining severity, volume saturation, representativeness, and evidence; and (4) evaluate a synthetic program of 240 reports. Prompt-level fixes were fastest (median 2.1-3.4 days) but less durable (21-38% recurrence); dataset edits and reward tweaks were slower (13.5 and 21.9 days) yet more durable (12-18% recurrence) with higher planner uptake (30-36%). A threshold of 0.12 yielded 93% precision and 75% recall; increasing representativeness raised recall to 81% with little precision loss. We discuss integration with participatory governance, risks (e.g., overfitting to vocal groups), and safeguards (dashboards, rotating juries).

Problem

• Addressing group-level harms in urban visualization diffusion models
• Developing collective visual bug reports for model and workflow fixes
• Establishing recourse mechanisms with severity and representativeness thresholds

Innovation

• Community visual bug reports trigger fixes
• Four recourse primitives within diffusion stack
• Mandate thresholds combining severity and representativeness