Akademik Veri Yönetim Sistemi
TÜRK DENETİMLİ SERBESTLİK SİSTEMİ 20. YIL ULUSLARARASI KONGRESİ, Antalya, Türkiye, 1 - 03 Aralık 2025, ss.111, (Özet Bildiri)
Modern probation systems increasingly prioritise the reduction of reoffending by foregrounding social reintegration and attending to the environmental and social contexts within which individuals return to community life. A substantial body of scholarship shows that crime is profoundly shaped by both physical and social environments (Bursik Jr, 1988; Cohen & Felson, 1979), and that community perceptions are closely coupled with the spatial distribution of crime (Abraham et al., 2025; Brantingham & Brantingham, 1981). This nexus between spatial perception and crime is especially salient for probation practice: when an area is perceived as unsafe, informal social control may be attenuated, thereby creating conditions conducive to offending; conversely, the same perceptions influence the degree of social acceptance extended to people under supervision. Advances in digital technology now make it feasible to map public perceptions at scale using big data and AI. Crowdsourced initiatives such as MIT’s Place Pulse 2.0 provide systematic measurements of urban perception across multiple dimensions, yielding intelligence valuable for spatial planning, risk assessment, and resource allocation within probation services. In Türkiye, integrating digitalisation and data-driven approaches into the probation system is therefore of particular strategic importance. In Istanbul’s Fatih district, we analysed 6,991 street-view images collected between 2014 and 2024 from 1,709 streets and avenues. Leveraging AI models trained on the Place Pulse 2 dataset (Dubey et al., 2016), implemented via the ZenSVI library (Ito et al., 2025), each image was evaluated across six perceptual dimensions: safety, wealth, beauty, boredom, depression, and liveliness. The results were mapped to produce district-wide perception surfaces, and inter-dimension correlations were computed to characterise co-variation among perceptions. Two distinct types of street or avenue profiles were identified. A majority of segments (66%) were characterised as “more liveable,” showing higher average ratings across multiple dimensions, including safety, affluence, aesthetic appeal, and vibrancy. The remaining 34% were deemed “less liveable,” with lower scores in these same areas. Notable positive correlations emerged between perceived wealth and vibrancy, as well as between aesthetic appeal and wealth. Negative associations were found between boredom and vibrancy, and between safety and feelings of depression. This study demonstrates the efficacy of crowdsourced perception data and AI techniques for producing evidence-based perception maps to inform probation practice. Spatial clusters with lower liveability scores align with areas likely to require enhanced supervision and strengthened reintegration support. The findings can directly shape service delivery by: (1) identifying high-risk neighbourhoods during re-entry; (2) guiding the allocation of community oversight resources in line with public perceptions; and (3) enabling the design of place-specific rehabilitation programmes. Future work will extend the methodology through contemporaneous community surveys and the training of a new AI model on the expanded dataset, thereby establishing a replicable framework that advances the digitalisation of probation for crime prevention.