Marked by Time

Marked by Time speaks to contemporary debates about criminal risk assessment and the manifestations of what are typically assumed to be our most inner traits. I show how shifting contexts—levels of violence, policing strategies, lead exposure, the rise of community-based organizations, technological conditions, and more—challenge the common belief in stable, individual propensities to commit crime and in timeless rules for predicting risk.

This matters because risk assessment pervades the criminal justice system and many other domains of life. Formal risk tools guide pretrial release and probation decisions, for example, and criminal-history information informs sentencing, employment and tenant screening, and occupational licensing. With the rise of artificial intelligence and large-scale databases, predictive risk assessment is accelerating, and recent legislation has called for still more tools.

I show the danger of treating these predictions as timeless. When I built prediction instruments using data from children in my study who were born in the 1980s—using familiar risk factors such as family poverty, low self-control, being male, and living with a single parent—those instruments overpredicted the arrest probability of children born not much later, in the mid-1990s, by more than 50 percent. The world had changed: crime declined, and the relationships between these risk factors and arrest shifted.

History, in other words, is baked into criminal records. In this case, it generates a systematic gap between actual and predicted arrest patterns across birth cohorts. This bias appears in all major racial groups and among youth who were arrested as juveniles, so it is not the usual form of bias we worry about. It also shows up for both absolute and relative risk. A prediction tool can correctly rank who is riskier than whom and still exaggerate how risky anyone actually is. As in medicine, what ultimately matters is absolute risk—the actual probability of harm.

Self-control offers a second example of the problem. Low self-control—the inability to regulate impulses—is widely studied and often assumed to be a stable predictor of crime. In my data, low self-control is indeed linked to arrest. But individuals from the 1980s cohort with high self-control had the same arrest rates at age 20 as individuals from the 1990s cohort with low self-control. The meaning of self-control, in terms of real-world outcomes like arrest, had changed over a relatively short time period. 

I came to these kinds of questions relatively late in my career. For decades I have studied life-course trajectories of crime and well-being, and the role of social contexts like neighborhoods. Only about ten years ago, when the youngest participants in our study were finally coming of age, did I fully grasp what our multi-cohort design—following several groups of children born in different years—could reveal about how social change transforms individual lives and undermines confident prediction. Marked by Time brings that realization to contemporary debates about risk, prediction, and responsibility.

Ongoing thread. More from Robert J. Sampson to follow.