Slashdot Mirror
AI is Sending People To Jail -- and Getting it Wrong (technologyreview.com)
At the Data for Black Lives conference last weekend, technologists, legal experts, and community activists snapped the kind of impact AI has on our lives into perspective with a discussion of America's criminal justice system. There, an algorithm can determine the trajectory of your life.
From a report: The US imprisons more people than any other country in the world. At the end of 2016, nearly 2.2 million adults were being held in prisons or jails, and an additional 4.5 million were in other correctional facilities. Put another way, 1 in 38 adult Americans was under some form of correctional supervision. The nightmarishness of this situation is one of the few issues that unite politicians on both sides of the aisle.
Under immense pressure to reduce prison numbers without risking a rise in crime, courtrooms across the US have turned to automated tools in attempts to shuffle defendants through the legal system as efficiently and safely as possible. This is where the AI part of our story begins. Police departments use predictive algorithms to strategize about where to send their ranks. Law enforcement agencies use face recognition systems to help identify suspects. These practices have garnered well-deserved scrutiny for whether they in fact improve safety or simply perpetuate existing inequities.
Researchers and civil rights advocates, for example, have repeatedly demonstrated that face recognition systems can fail spectacularly, particularly for dark-skinned individuals -- even mistaking members of Congress for convicted criminals. But the most controversial tool by far comes after police have made an arrest. Say hello to criminal risk assessment algorithms.
Under immense pressure to reduce prison numbers without risking a rise in crime, courtrooms across the US have turned to automated tools in attempts to shuffle defendants through the legal system as efficiently and safely as possible. This is where the AI part of our story begins. Police departments use predictive algorithms to strategize about where to send their ranks. Law enforcement agencies use face recognition systems to help identify suspects. These practices have garnered well-deserved scrutiny for whether they in fact improve safety or simply perpetuate existing inequities.
Researchers and civil rights advocates, for example, have repeatedly demonstrated that face recognition systems can fail spectacularly, particularly for dark-skinned individuals -- even mistaking members of Congress for convicted criminals. But the most controversial tool by far comes after police have made an arrest. Say hello to criminal risk assessment algorithms.
I have been at the receiving end of these. They are far simpler and more biased than anyone would reasonably expect. They are likely the result of a self-styled 'expert', as they are certainly never vetted in any academic sense, and they show a lot of that person/teams biases. They put the vast majority of the weighting on income and upbringing, meaning a minor non-violent criminal with a background of childhood poverty will be treated as much larger threat than a wealthy murderer. Just for a bit of background, I shot someone. I have no delusions that the money and privilege of being a white professional with a lawyer changed my outcome in the courtroom and DA negotiation side. Were I poor and minority I likely would have been in the system for life. But beyond that, once into the system, the imbalance continued. On a threat scale of 0 to 100, I rated something like a 2. It was absurd. Mainly because the questions were weighted toward things like how long I had been at an address, and if I owned or rented. Also previous convictions REGARDLESS OF TYPE. I knocked all those questions out of the park. If some poor minority kid with little education and a few tickets and a minimum wage job who had recently changed apartments got caught with a joint, they would have scored something like at least a 50 as a baseline. Which means I got an immediate and almost unsupervised walk (not a day in jail, call in once a month), while our hypothetical joint owner would have been locked up with at least a medium threat rating. TL,DR; This has nothing to do with AI, more what some white, educated social worker pulled from very flawed data filtered through their biases.
The US is less tolerant of crime and better at catching it than other countries. That's why we have more people in prison. Black people commit more crimes so they go to jail more often. Those are facts. Get over it. There isn't some stupid national conspiracy to put non-white people in prison more. They just simply commit more crimes period. Fix it or shut up about it all being "the system's" fault.
Quoting the Slashdot story summary: "The US imprisons more people than any other country in the world."
It is more correct to say, "The US imprisons a higher percentage of its people than any other country in the world."
Either is correct. America is the world leader both in percent and absolute number of prisoners. China is the only other country that even comes close. China imprisons about a quarter as many people as a percentage, but even if you include the "re-education" camps in Xinjiang they are still below America in absolute numbers.
Prison is a big, profitable business in the United States. The companies that manage prisons are paid up to $70,000 per prisoner, per year.
Private prisons are a problem, and in my opinion should be shut down. But prison unions in government run prisons are also a big problem. The California prison union was a big financial supporter of the "three strikes" law that caused prison populations to soar, locking up thousands of non-violent geriatric geezers that should be in nursing homes instead of prison cells.
Private prisons and prison unions both work to not only lock up more offenders and lengthen sentences, but also to increase recidivism. It is well known that prisoners that keep in touch with their families and friends are more likely to successfully reintegrate with society. So the prisons actively work to prevent that, by moving prisoners out-of-state, denying visits for capricious reasons, and making phone calls expensive and infrequent.
It is a rotten corrupt process, and we all pay the price.
no, the point of the article is that algorithm training usually carries the implicit assumption that the data used to train is good, in all important ways, but the history of law enforcement has not always been fair, and that causes problems with the algorithm's outputs.
Actually, one of the defining characteristics of psychopathy is an undersensitivity to punishment or the threat of punishment. Psychpaths do have a high degree of goal orientation. There's some interesting programs for treating psychopathic youths by teaching them to find rewards within the boundaries of social norms.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.