AI-Assisted Participatory Design System for Urban Safety Enhancement Based on Crime Prevention Through Environmental Design

Abstract This study proposes an artificial intelligence (AI)-assisted participatory design support system for crime prevention through environmental design (CPTED) in low-rise residential alleyways. Practitioners have widely developed CPTED guidelines, but they often rely on representative cases and expert interpretation, which limits their adaptability to diverse, context-dependent everyday spaces. To address this limitation, we integrate image-based spatial analysis with generative AI to support user-centered CPTED design exploration. Users upload photographs of alleyways, and the system automatically analyzes the images to extract object- and material-level spatial features relevant to CPTED interpretation. The system then translates these analytical outputs into structured semantic conditions that preserve the original spatial context while guiding CPTED-oriented visual interventions. Based on these conditions, a Stable Diffusion-based image-to-image pipeline generates visual design alternatives that reflect CPTED strategies without substantially altering the underlying spatial configuration. The system further incorporates an iterative feedback mechanism that allows users to provide natural language input, which the model progressively reflects in regenerated design proposals. Preliminary evaluation through a resident survey (N = 200) and an expert questionnaire (N = 17) suggests that general residents perceive the system’s outputs as meaningful safety improvements and that the rule-based pipeline produces outputs rated as more appropriate from a CPTED perspective than non-rule-mediated alternatives. By enabling residents to engage directly with visualized CPTED improvements and iteratively refine design outcomes, the proposed system extends CPTED design from an expert-driven process to a participatory and context-responsive workflow. This study suggests the potential of generative AI as a visualization tool and a mediating mechanism that connects spatial analysis, design decision-making, and user perception. This approach offers a flexible, context-responsive framework for safety-oriented design in irregular urban environments.