Algorithmic risk assessments and the double-edged sword of youth.
By M. Stevenson, C. Slobogin. 2018
27 citations recorded by the source corpus.
At sentencing, youth can be considered both a mitigating circumstance because of its association with diminished culpability, and an aggravating circumstance because of its association with crime risk. In theory, judges and parole boards can recognize this double-edged sword phenomenon and balance the mitigating and aggravating effects of youth. But when sentencing authorities rely on algorithmic risk assessments, a practice that is becoming increasingly common, this balancing process may never take place. Algorithmic risk assessments often place heavy weights on age in a manner that is not fully transparent - or, in the case of proprietary "black box" algorithms, not transparent at all. For instance, our analysis of one of the leading black-box tools, the COMPAS Violent Recidivism Risk Score, shows that roughly 60% of the risk score it produces is attributable to age. We argue that this type of fact must be disclosed to sentencing authorities in an easily interpretable manner so that they understand the role an offender's age plays in the risk calculation. Failing to reveal that a stigmatic label such as "high risk of violent crime" is due primarily to a defendant's young age could lead to improper condemnation of a youthful offender, especially given the close association between risk labels and perceptions of character and moral blameworthiness.
Subfields: Clinical Psychology, Safety Research, Sociology and Political Science
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Clinical Psychology
In Safety Research
In Sociology and Political Science
Citations in the corpus
- The age of secrecy and unfairness in recidivism prediction (2018, background): “However, their analysis has similar issues to that of ProPublica or (29): all of these groups attempted to explain COMPAS scores using ProPublica’s limited set of features, but this type of analysis is not valid because there are many unmeasured features.”