On The Docket, a new independent project to expand access to the Supreme Court, is using AI to generate visual depictions of U.S. Supreme Court justices reading their decisions.
On The Docket
hide caption
toggle caption
On The Docket
Artificial intelligence, meet the U.S. Supreme Court.
It’s an institution steeped in tradition and resistant to any quick changes in the way it does things. But like it or not, the justices are about to see artificially created versions of themselves, essentially avatars, speaking words that they actually did speak in court but that were not heard contemporaneously by anyone except the people in the courtroom.
Northwestern University professor Jerry Goldman has been creating ways to make the Supreme Court more accessible to the public since 1996 when his nonprofit project, Oyez, went live on the internet. The site sought to provide audio of the court’s oral arguments and opinion announcements in every case decided by the Supreme Court dating back to 1955 when the court began taping its courtroom proceedings.
Professor Goldman’s Oyez project was a huge deal when it debuted because until the early 1990s, the public had no idea the court had been taping its courtroom sessions. And the process for preserving the tapes was so helter-skelter that many of those recordings were lost for all time. What’s more, access to the audio was severely limited. Indeed, nobody outside the court had access until months and months after the case was heard and decided. Only at the beginning of the next court term were the audio tapes generally made available from the prior term.
It wasn’t until 2020 when COVID-19 shut down the country that the court was essentially forced to allow every oral argument to be broadcast live, with the justices linked by phone line and the public able to listen in. And after the pandemic, the justices, who had long resisted regular audio broadcast of their arguments, without fanfare, left the pandemic system in place.
That has left just one part of the court’s public business unavailable on a same day basis. The all-important announcements of decisions, summarized by the justices from the bench, as well as, on occasion, oral dissents. To this day, the old system of limiting access until the following term remains intact, so that only those people actually in the courtroom can hear and see the drama of the day.
Now Goldman’s team is experimenting with making that drama ever more real, even though the audio remains unavailable for months. They are using AI to recreate what people in the courtroom not only heard when decisions were announced, but what they saw.
As Goldman puts it, “Since it’s public in the courtroom, it should be public for everybody. That’s simple.”
But with no cameras allowed in the Supreme Court, how is Goldman’s new site, On The Docket, creating the visuals?
Answer: essentially with avatars. And it wasn’t easy, says University of Minnesota professor Timothy R. Johnson, one of the architects of the project, along with Spooler, an AI design company. He says some of the initial AI efforts were hilarious.
“We have bloopers where they would give the robot a particular command and it would do something completely uncanny,” like having certain justices magically disappear on the bench, or having all of them bend forward at the same time.
In the end, by using photos and videos from the justices’ public appearances, they were able to create real-looking video versions of each justice and things like their mannerisms, head tilts and hand gestures. They were able to match these avatars with the existing real audio.
There were, of course, ethical questions they had to tackle. Should they make the video look completely real, or do something special to tell the viewer it is not. In the end they opted for slightly cartoonizing the video, and clearly marking it AI generated so viewers would know what was real —the audio — and what was not —the video.
In their first foray into this brave new world, for instance, they have created a visual of Chief Justice John Roberts’ 14-minute summary from the bench of the court’s 6-to-3 decision granting then former President Trump, and all future former presidents, complete immunity from prosecution for any core official actions taken while in office, no matter how nefarious.
After Roberts is Justice Sonia Sotomayor, also on the bench, outlining her dissent.
Together, their passionate spoken words, lasting a total of 38 minutes, are both riveting and a bit eerie.
The court is probably not going to like this latest rendering of reality. After all, it kept the taping of oral arguments and announcements secret until 1993 when law professor Peter Irons signed the pledge not to disclose the then-secret recordings and then published the oral arguments in a book that included dubbed cassettes of arguments he thought very important.
The court promptly sued him, but soon dropped the case, apparently having concluded it was a loser.
Since then the oral arguments are routinely broadcast, thanks to the COVID-19 pandemic, but the fascinating bench announcements of court decisions have remained under lock and key until many months after the decisions are rendered. Reporters have repeatedly asked that the opinion announcements be made available for live broadcast, and scholars have made similar requests.
Professor Goldman, for example, notes that he has reviewed papers from the early Warren Court in the 1950s in which the justices discussed memorializing the oral arguments and opinion announcements with tape recordings.
“There is no indication in those papers that they wanted to keep these secret,” he says.
As far as is known, however, every request from reporters or scholars that the court make the bench announcements accessible for live audio broadcast has been met with deafening silence. And even AI can’t capture that for audio or video.
