Uma arquitetura de sensoriamento social para suporte à comunicação não planejada no contexto de violência urbana no projeto CidadeSocial

Abstract

Violence in large urban centers poses significant challenges for both citizens and institutions, requiring public administrators to manage official information dissemination while residents seek ways to access critical data beyond their immediate social networks. The widespread use of social media has enabled individuals to act as human sensors, reporting urban violence through mobile devices. This study proposes and evaluates the feasibility of a software architecture to support unplanned communications and automatically detect/classify violence-related messages by severity level, adopting a Design Science Research methodology that includes: (1) problem context analysis and literature review on automated violence detection, (2) conception of a social sensing-based architecture, and (3) experimental validation through human sensor emulation, machine learning model development for violence classification, and architecture testing with multiple messages. Results demonstrate the viability of social sensing for enabling communication among unconnected individuals and confirm the effectiveness of human-as-sensor approaches for violence detection, offering practical solutions for urban safety challenges.