More than 200 court rulings in 2024 cite an algorithm that classifies foreign prisoners without ties as high risk, even if they lack other risk factors
In 1993, two beekeepers stumbled upon a gruesome discovery in a ditch near the La Romana ravine in Valencia: the graves of three teenage girls Toñi, Miriam, and Desiré, whose bodies began to surface after a night of torrential rain. What followed was a national reckoning. The Spanish Civil Guard quickly uncovered the extent of the crime: these girls had been brutally tortured, raped and murdered. Known now as the Alcàsser Murders, the case transfixed Spain; its lurid details and courtroom drama unfolding nightly on television. The trauma endures: a recent Netflix documentary revived the unease, reintroducing the crime to a global audience.
But the legacy of the Alcàsser case reaches beyond collective memory. In a recent exposé for Civio, journalist Ter García traces how the crime catalyzed a new form of state logic: the creation of an algorithm to assess the risk of prison leave. Known as the Table of Risk Variables (TVR), the tool, introduced in the mid-1990s, was designed to quantify the probability that an inmate might violate the terms of their exit permit. “In those years, no one was doing that,” says José Ángel Brandariz, a criminal law scholar.
“The context was very pressing,” recalls Clemente, a legal psychology expert from the University of A Coruña. One of the perpetrators of the Alcàsser Murders had escaped from prison the previous year using a prison permit. In the aftermath of the murders, public anxiety soared. The TVR offered a bureaucratic balm, a promise that science and objectivity could replace fallible human judgement in determining who could be safely released, and when.
Yet, thirty-two years later, the same formula remains unchanged. García reports that in 2024 alone the TVR was cited in more than 200 rulings to override the recommendation of prison officials and deny prison release permits. “It’s a system that’s preventing more inmates from going out onto the streets to avoid problems, even at the cost of their rehabilitation,” Clemente warns.
Critics of the algorithm argue that its foundations are flawed. Margarita Aguilera, an advocate for incarcerated women, notes that the model’s original data sample consisted of just 1,500 inmates, of which only 62 were women. “It’s based on a population that bears no relation to today’s”, she says, adding that the reasons why a woman might violate a permit can differ profoundly from those of men.
Foreign nationals fare worst under the TVR. Non-EU prisoners are automatically assigned the highest risk, regardless of their behaviour or other metrics. “Foreigners are screwed,” says Carlos García Cataño. “And even if you have legal roots, if you’re serving time outside your home religion, they won’t give you a permit either”.
Perhaps most troubling of all is the algorithm’s middling accuracy. According to García’s research, the TVR correctly predicts permit violations only 53.33% to 69.69% of the time, a coin toss in the bureaucratic frame.
Calls for reform have gone largely unanswered. Clemente, who once hoped the algorithm might evolve, now seems resigned. “To this day I haven’t heard anything,” he says: “which I find serious.”
Original article (“Las prisiones españolas usan un algoritmo sin actualizar desde 1993 para decidir sobre permisos de salida”) by Ter García was first published in Spanish on 26 February 2025.
It’s available here.
Civio is the first organisation in Spain to specialise in monitoring public authorities, which they achieve through data journalism. They call for transparent governments and institutions.
