Poster #42 - Recidivism Poles in Motion: Integrating Multidimensional Scaling and Survival Analysis for Youth Recidivism Risk

• In Event: Poster Session I (posters 1 - 79)

Abstract

Risk assessment tools such as the Youth Level of Service/Case Management Inventory (YLS/CMI; Hoge & Andrews, 2002) provide structured evaluations of recidivism risk across multiple domains. However, conventional analyses typically rely on single-time-point summaries of recidivism risk, offering a static perspective that fails to capture how the relational structure among risk factors evolves as events unfold. We introduce a spatial visualization framework that integrates classical multidimensional scaling (CMDS) with survival analysis by representing survival analysis–based quantities as reference points (“poles”) within a CMDS configuration. Specifically, we define a recidivism pole based on null-model martingale residuals and a non-recidivism pole based on Kaplan–Meier pseudo-values. By advancing the evaluation time τ, we track pole migration—the movement of these reference points through the risk factor space as events accumulate and the risk set evolves. This dynamic representation reveals how clusters of risk factors reorganize over time and how their configuration relates to recidivism outcomes. The proposed framework offers a novel perspective on the temporal dynamics of recidivism risk and provides a foundation for more interpretable, time-sensitive risk assessment.

Authors

• Takemi Mori, Konan Women's University

• Yoshikazu Shimada, Fukuoka Family Court

• Yoshimasa Matsuda, Fukuoka Prison