Listen to the episode (in spanish)
For more than a decade, prisons in Catalonia have been using an algorithm on their inmates. This algorithm, developed in 2008 is called RisCanvi. It is a tool created by the Group of Advanced Studies of Violence, led by Antonio Andrés Pueyo, and its main purpose is to predict whether the person incarcerated is likely to reoffend.
RisCanvi has evolved over the years and today it is also used to estimate whether the inmate is a threat to themselves, other inmates or if they are at risk of violating probation. This tool uses past and present data of the person it analyses. What has their conduct been during their sentence? What personal responsibilities does the inmate have outside of prison? Does the offender suffer from drug or alcohol abuse?
Although the internal workings of RisCanvi is complex, the algorithm yields a simple result. The system assigns prisoners a level of risk: low, medium or high. This information is later used by judges to help them decide whether to grant or deny parole.
In this episode, (available only in Spanish) we will hear a first-hand account by Andrés Pueyo as he explains the creation and workings of this algorithm. We will also visit Quatre Camins, a prison in Barcelona and meet prison manager, Griselda Baldris, as she recounts how RisCanvi is used inside the prison.
The use of predictive algorithms in penal and penitentiary systems has generated hot debate among experts. This is the first episode of a two-part series. In the following episode we will take a deeper look and ask… can an algorithm really predict criminal recidivism?