Assignment: Spain - can an algorithm predict murder?

Early on a Sunday morning in February in the Spanish seaside town of Benalmadena, Catalina, a 48-year-old mother of four, was killed at home – the building was set on fire. Her ex-partner was arrested and remains in custody. In January, Lina – as she was known to her family and friends – had reported her ex-partner to the police for ill-treatment and threatening behaviour. And by doing so, she became one of around 100,000 cases of gender-based violence active in Spain’s VioGen system.

VioGen is an algorithm used by the police – it’s a risk assessment tool. Based on a woman’s answers to a series of questions, it calculates the likelihood she will be attacked again so police resources can be allocated to protect those most in danger. The level of risk could be negligible, low, medium, high or extreme. Lina was recorded as being at ‘medium’ risk of a further attack by the man who was her ex-partner. Three weeks later, she was dead. VioGen’s critics are concerned about the number of women registered on the system who are then murdered by men who are former or current partners. Its champions claim that without VioGen there would be far more violence against women.

With AI in the ascendency, and governments increasingly turning to algorithms to make decisions affecting society, for Crossing Continents, Linda Pressly and Esperanza Escribano investigate the story of VioGen and domestic violence in Spain.