FUZZY RNR: A MAMDANI DECISION SUPPORT SYSTEM FOR PRISONER ASSESSMENT AND REHABILITATION IN CENTRAL JAIL KHUZDAR

Abstract

Smart prisoner assessment and rehabilitation planning are key to public safety and successful reintegration. Standard assessment processes tend to have difficulties with complexity, subjectivity, and subtle interplay between criminogenic and responsivity factors, soft interventions that are not necessarily optimally adjusted to the individual. This study designs and tests a Mamdani fuzzy logic decision support system for holistic prisoner assessment and rehabilitation program recommendation based on the Risk-Need-Responsivity (RNR) model. The system combines several attributes, such as crime severity, psychological effect, repeat-offender status, educational level, need for substance misuse, need for mental health, social support, work history, motivation, cultural variables, and in-prison conduct, and represents them using Gaussian and triangular membership functions to portray inherent imprecision. A rule-based inference engine, based on criminological literature and expert opinion, controls the inference engine. This study evaluated the system using a dataset collected from the Central Jail in Khuzdar. Results show robust categorization across RNR profiles (low/medium/high risk, need, responsivity) and context-sensitive recommendations for rehabilitation intensity (none, low, moderate, intensive). Distributional and sensitivity analyses confirm stability and highlight rules with the largest influence on outcomes. The resulting tool is transparent, interpretable, and adaptable, and can help correctional professionals assign more precise, culturally sensitive interventions in resource-constrained correctional settings.