Imagine that you have been convicted of a crime and an algorithm is to help the judge by proposing the sentence. When computers are programmed, discrimination is built into the algorithms, so if you look a certain way you get a harder sentence. This example is not fiction. It comes from the USA where AI is being used in the legal system to propose sentencing for criminal offences, and it proposes harder sentences for black people. We don’t know why. It might be chance, a prejudiced system developer or perhaps distorted data that the system had to practice on.
“As human beings, we must be aware that every day we are subject to decisions made by an algorithm. There is a risk that we give machines the power to take decisions about us even though we do not fully understand how the underlying technology works,” says Stanley Greenstein senior lecturer in legal informatics.

Crucial everyday decisions

Stanley Greenstein likens algorithms to black boxes. We know what we feed into the box and what comes out, but sometimes even the people who develop them don’t know exactly how the algorithms work. They are self-teaching, can be altered in real time and are very complex. They are already taking decisions about us every day. They predict the music you wish to listen to and the films you want to see, but they can also decide how large a mortgage the bank gives you or whether you get a job.

Stanley Greenstein is researching how the law should relate to artificial intelligence. How can the algorithms that make decisions become more transparent and how can we safeguard the rule of law. So far he cannot give any simple answers, but he believes that it is a good beginning if we talk about it and don’t close our eyes to the risks. These are complex issues. The law protects the system through intellectual property law, for example, and the algorithms are beginning to be too advanced for our cognitive abilities to keep up with.
“We people will not be able to monitor the technology. I believe the answer lies in technology. It is a basic principle of the subject of legal informatics to proactively build the law into the technology from the beginning. Perhaps we also need a certification system or an authority to scrutinise algorithms,” he says.

Room to test new ideas

Stanley Greenstein has received project funding through the digital human science initiative which has resulted in a chapter in the anthology to be published by Stockholm University Press next year. His research does not begin or end there. Legal informatics is by definition interdisciplinary, but he welcomes the opportunity for meetings that the initiative gives, as well as the room to test new ideas.
“Sometimes we can find ourselves in a bubble together with others who think the same way. Meeting people from other academic disciplines gives distance and new perspectives. Tomorrow’s complex problems in the digital society will need interdisciplinary solutions.”