On February 12, the Ministry of Justice announced plans to use predictive policing to overhaul the youth justice system. Hidden in the 25-page document was a proposal to use “machine learning and advanced analytics” to “support early and appropriate intervention” in youth crime.
The white paper was vague on details, only promising more news in the spring. times The article explained the plan in more detail. Under an inflammatory headline promising a machine that predicts “future criminals,” the column explained:
Artificial intelligence (AI) could be used to predict future criminals under a government plan to identify children who need targeted intervention to stop them falling into a life of crime. […]
Academic research suggests patterns may emerge from data collected by health visitors checking on newborns, but it is unclear whether government programs will go back that far to determine whether someone is at risk.
Now, it would be easy to point out here that this pre-crime policing is horribly dystopian. It is a phrenological measurement of the skull, Minority Report.
And it’s true, it’s a horrifying dystopia. But this is also the current reality that racist people in Britain have been exposed to for decades.
Predictive policing and “criminals of the future”
Regarding the AI plan, government officials said:
We’re looking at how we can use AI and machine learning more effectively to essentially predict future criminals, ethically and morally. It’s about ensuring that data from the NHS, social services, the police, the Department for Work and Pensions and the Department for Education is used effectively and that AI can be used to achieve more than we can currently do.
This will revolutionize the way we invest money and resources in prevention. We continue to get the same profile of offenders in our justice system, but intervention is too slow.
This isn’t about making people criminals, it’s about enabling them to better understand the alarms in their systems and do better with data and AI modeling.
Jake Richards, Minister for Youth and Justice, further explained:
I am determined to harness the power of artificial intelligence and machine learning to gain better insight into the root causes of crime. This allows us to focus on early intervention for individuals and families, delivering better outcomes for children and keeping our communities safer.
However, we must hold and use this personal data carefully. That’s why I asked this expert panel to examine not only the effectiveness but also the ethical and legal implications of this study.
of times Additionally, it says data shows that neurodiverse, poor and ethnic minority children are more likely to commit crimes. Four out of five children in youth detention are neurodivergent. 33% of children from care backgrounds receive a warning from the police before they turn 18.
This article says all this in a neutral tone that only a newspaper chosen by discerning bigots could manage. And, of course, that is a seriously misleading misuse of the truth.
The past is biased, and the future is also biased.
In fact, these marginalized children are more likely to be arrested, cautioned, and prosecuted by the police. Police create a profile of the person arrested. They have (racist and discriminatory) ideas about who criminals are and police people accordingly. And surprisingly, people who were treated as criminals were being arrested one after another.
That is completely different from “having a high probability of committing a crime.”
AI decision-making is sometimes seen as unbiased and emotionless, but this is far from the truth. Rather, it simply hides the very human biases in the training dataset behind a veneer of ruthless “fairness.”
Ashwini KP, the United Nations Special Rapporteur on modern forms of racism, specifically called for predictive policing in her report on AI bias in policing. In 2024, Ashwini explained:
Predictive policing can exacerbate historic over-policing of communities along racial and ethnic lines. Because law enforcement officials have historically focused on such neighborhoods, police records overrepresent community members from these neighborhoods. This influences where the algorithm predicts future crimes will occur, leading to increased police deployment in problem areas. […]
As officers in heavily policed areas record new crimes, a feedback loop is created where the algorithm generates increasingly biased predictions targeted at these areas. In other words, past bias leads to future bias.
criminalization before crime
However, as mentioned earlier, this feedback loop is not a problem unique to AI itself. Rather, it is inherent in the very idea of pre-crime policing, an oppression that racist people in Britain have been dealing with for decades.
Take, for example, the Metropolitan Police’s Operation Trident in the 1990s. This was an attempt to stop gang-related violence in London, but instead resulted in large-scale racial profiling against young black people. Amnesty International’s report on Trident’s Gang Matrix database states:
The types of data collection that underpin the Gang Matrix disproportionately focus law enforcement efforts on Black boys and youth. It erodes their right to privacy based on what may just be their peers in the area they grew up in and how they represent their subculture in music videos and social media posts. Officers from the borough’s gang control unit monitor the social media pages and online interactions of people deemed “at risk” of gang involvement, invading the privacy of far more people than those involved in any type of misconduct.
Then, in 2003, the UK government developed the Prevent Counter-Terrorism Strategy. Ostensibly, it aims to prevent people from becoming radicalized into extremist ideologies. In practice, it unfairly targets Muslims, including Muslim children, as a dangerous “other” for surveillance and hostile treatment.
And in 2023, the Shawcross Review of Prevent baselessly argued that the strategy should further target Muslims rather than far-right extremists. As such, this was the perfect epitome of prejudice confirmation in action. at that time, Canary-san Mariam Jameela writes:
Pre-crime strategies like Prevent always assume full agency and power for all Muslims. For that to happen, there must be a cultural belief that Muslims are always potential terrorists and therefore objects of suspicion. Underlying this presumption is that there is something sinister about Islam itself. Over the years, successive governments have created a culture of criminalization that sees Muslims as nothing more than potential criminals.
Now, and Jake Richards’ protest that his AI plans use data ethically to produce better outcomes for children certainly sounds like the same kind of discriminatory leftovers. We have already seen what the ethics and care of these people are.
There is no way, machine learning or otherwise, to predict crime that is not driven by our prior biases. All this “new” strategy can do is push further marginalized youth into pre-criminal no-man’s land. And all the while, vulnerable children will be told firsthand that their every move is under constant surveillance.
Featured image via Canary
