Artificial selection

Building an Artificial Intelligence “Good Stuff In, Good Stuff Out”

Garbage in, garbage out is the cliché that drives most conversations about AI. While people have generally come to recognize AI as a valuable tool for limited – albeit essential – legal tasks, fear of a GIGO outcome still haunts the region. Artificial intelligence has made huge strides in discovery, brief writing, diligence and contract management, we have also seen retreats in areas where the tool can so easily become your racist robot uncle, such as facial recognition.

But the potential that AI could bring to justice remains too great to simply throw in the towel.

A few weeks ago I spoke with Neil Sahota, an IBM Master Inventor, United Nations AI expert, and lecturer at UC Irvine on building better AI. I had planned to write this article earlier but – I don’t know if anyone noticed – the Supreme Court decided to resume the legal news cycle for a week or four.

Anyway, Sahota told me that one of the biggest areas of interest for the UN (especially UNICRIthe United Nations Interregional Crime and Justice Research Institute) involves the development of AI judgment, which would certainly require guarantees that good things come in.

Having AI judges is anathema to most American audiences (But see the last four weeks). If robot lawyers are scary, robot judges register as completely terrifying. Even with the bias inherent in the legal system, the experience of bias-amplifying algorithms strikes the American ear as, at best, a lateral move.

However, it speaks to the American privilege of a generally strong respect for the rule of law (But see Dobbs v. Jackson Women’s Health Organization, 597 United States ____ (2022)). Sahota explained that many national legal systems are rampantly corrupt with biases showing up in even the most mundane legal fights. “People here don’t think the system is corrupt – bias sure, but people don’t think they’re getting shot on a bunch of bogus charges,” he told me.

A reliable artificial intelligence lawyer capable of — dare we say it? – Dealing with single business bullets and strikes potentially revolutionizes the rule of law in some parts of the world. Consider Estonia moving the traffic court to AI with an appeal procedure for outlier cases. Which is another dimension of the international problem: many countries have people waiting years to solve small problems. Automated systems can dispense with routine cases by easily freeing up human intervention for more complex issues.

Closer to home, Sahota said AI can play a better role in jury selection and evidence assessment if properly trained in psychography and neurolinguistics. Remove unconscious bias from jury questionnaires? It’s not the answer to a specific question – most people won’t tell you right away – it’s in the patterns that develop.

“Language is like a fingerprint. Especially the more you speak and write,” Sahota explained. Beyond detecting juror bias, it provided a stunning example of AI breaking down deposition testimony. Taking the artificial intelligence tool into an already concluded case of abuse of the Catholic Church, the product compiled hours of testimony from a cardinal and immediately grasped the fact that he called other priests ‘friends’ or ‘colleagues’, with the exception of seven whom he consistently identified as ‘fellows’. The AI ​​said it was suspicious – it turned out that these seven priests were ultimately guilty. This is the kind of power and insight that AI can bring if developed correctly.

“The interesting irony here is that people thought there was no way a machine could rate a human better than another human. But AI can study all that [statements, demographics, body language] in real time and provide instant feedback. It doesn’t take much to miss a subtle clue.

“Technology, like all tools, can be used both ways. It’s just about how we choose to use it. And using it properly requires a vigilant assessment of inherent and unconscious biases and how they can find their way into algorithms, which Sahota identifies as constantly in the minds of the people he works with.

So don’t give up on the AI ​​just because it spits out trash – work harder to take the trash out.


Head shotJoe Patrice is an editor at Above the Law and co-host of Think like a lawyer. Feel free to email tips, questions or comments. Follow him on Twitter if you’re interested in law, politics, and a healthy dose of college sports news. Joe also serves as Managing Director at RPN Executive Search.