Scientists at Carnegie Mellon University (CMU), Harvard University, and the University of Pennsylvania suggest cohort bias driven by societal change undermines risk assessment instruments (RAIs) used to evaluate crime likelihood.
The researchers examined criminal histories of individuals in Chicago over 25 years, determining a machine learning tool forecasting the probability of arrest between ages 17 and 24 for cohorts born in the 1980s overpredicted that likelihood for cohorts born in the mid-1990s by up to 89%.
They also found substantial cohort bias within racial-ethnic groups, which persisted even when including arrest measures from immediately before the ages for anticipated arrest and when limited to high-risk individuals.
CMU's Erika Montana explained, "Our findings show that the relations between risk factors and future arrest are not stable over time. As a result, prediction models that rely on these risk factors are prone to systematic and substantial error."
From Carnegie Mellon University Heinz College
View Full Article
Abstracts Copyright © 2023 SmithBucklin, Washington, DC, USA
No entries found