Title: The Dynamic and Endogenous Behavior of Re-Offense Risk: An Agent-Based Simulation Study of Treatment Allocation in Incarceration Diversion Programs

Abstract:

Although incarceration-diversion treatment programs are designed to facilitate societal reintegration and curb recidivism, their limited capacity compels policymakers to make prioritization choices that frequently depend on risk assessment instruments. Although these instruments offer predictive value, they generally conceptualize risk as a fixed, individual characteristic. This approach fails to account for the temporal evolution of risk or the way treatment decisions influence outcomes via social interactions. To address this gap, this study introduces a novel framework that conceptualizes reoffending risk as a human-system interaction, thereby connecting individual behaviors with system-level dynamics and endogenous community feedback. By employing an agent-based simulation calibrated with U.S. probation data, we assess treatment allocation strategies under varying capacity limitations and incarceration scenarios. Our findings indicate that no single prioritization strategy is universally superior. Rather, the efficacy of a policy is contingent upon specific temporal windows and system parameters: prioritizing low-risk individuals yields better results when long-term trajectories are the primary concern, whereas prioritizing high-risk individuals proves more effective in the short term or when incarceration results in abbreviated monitoring periods. These insights underscore the necessity of evaluating risk-based decision-making frameworks as sociotechnical systems characterized by long-term accountability, rather than treating them as disconnected predictive tools.

Source: arXiv Generated at: 2026-06-03 00:00:00 UTC