On fairness, algorithms and the law

Stuart Corner

“Fair is foul, and foul is fair,” chanted the witches in Macbeth, meaning that appearances can be deceptive: things can differ from how they appear.

And our idea of fairness, under the law, can fall foul of algorithms when computer scientists try to develop fair computer systems that make fair decisions, according to Mireille Hildebrandt, a research professor in interfacing law and technology at the Free University of Brussels.

She discussed the many and complex issues thrown up by the nexus of computer science and law in conversation with UNSW cybersecurity expert Richard Buckland, in a webinar New technologies and the law, co-presented by the UNSW Centre for Ideas and the Australian Society for Computers and Law.

TechLaw: Fairness in a world of algorithms and the law

Fairness, said Dr Hildebrandt, is now an important subdomain of computer science, as computer scientists strive to develop fair computing systems that make fair decisions. She said there was a big difference between how computer scientists and lawyers deal with the concept of fairness.

21 ways to be fair

“Computer scientists demonstrated a couple of years ago that you can define fairness in at least 21 different ways. That means when you implement [a] definition of fairness into the system, it will start taking decisions that are fair in that particular sense,” she said.

Lawyers, on the other hand “are very used to juggling with different fairness, conceptions and different perspectives on fairness. A lawyer is used to having … different considerations around fairness in their head, not always deliberately, a lot of this is tacit knowledge.

“But it’s a very different thing from what a computer scientist might want to do. They might be more interested in developing a universal, a general understanding of fairness.”

This, Dr Hildebrandt said, was just one result of a fundamental difference in the way lawyers and computer scientists approach problems: computer science is driven by data, ultimately represented as ones and zeros, and lawyers deal with text.

“Lawyers don’t talk about their study, their science and their practice being text driven, because it is so obvious… It’s like the water that a fish swims in, but it’s crucial.”

Computers, on the other hand depend on disambiguation. “If you need to write code … you have to be clear to the system, which meaning you want to formalise. You’re working with one or zero, and that trickles through the whole system, up to more abstract levels into computing systems.

“This means that you’re always starting out with mutually exclusive variables. Natural language is intuitively a very different sort of thing.”

Law, she argues, relies on natural language and “is all about ambiguity … [which] opens a space for contestation for different types of interpretation of the same word, the same sentence, paragraph or larger text body.”

Bringing computing and law together

To address these thorny issues Dr Hildebrandt said a new publication The Journal of Cross-Disciplinary Research in Computational Law is being launched. It is an open access journal and the first issue is due out in the second half of 2021. However, articles going into that edition are already available online.

“We have tried to bring together excellent papers in computer science and in law,” Dr Hildebrandt said. “Every paper is replied to by somebody from the other discipline, which means we hope to get a real conversation going.”

She argues that, although computer science and law are very different things, for many different reasons they are also “a match made in Heaven.”

She is also arguing for law to feature in computer science education, with an article Understanding law and the rule of law A plea to augment CS curricula, in a forthcoming edition of Communications of the ACM.

She has been awarded and European Research Council grant for a project Counting as a Human Being in the Era of Computational Law and says much of her research is now focused on the integration of artificial intelligence, but also things like introducing rules-as-code into the law.

She has also recently published a book, Law For Computer Scientists and Other Folk, which she describes “a result of eight years teaching law to masters students of computer science, who’s agile and inquiring minds made each course an intellectual feast.”

Do you know more? Contact James Riley via Email.

Leave a Comment

Related stories