Arizona’s Supreme Court is now using AI-generated reporters. Here’s why

Arizona’s highest court has created a pair of AI-generated avatars to deliver news of every ruling issued by the justices, marking what is believed to be the first example in the U.S. of a state court system tapping artificial intelligence to build more humanlike characters to connect with the public.

A court in Florida uses an animated chatbot to help visitors navigate its website, but the Arizona Supreme Court is charting new territory with the creation of Victoria and Daniel. Made of pixels, the two avatars have a different job in that they serve as the face of news coming from the court just as a spokesperson made of flesh and blood would do—but faster.

The use of AI has touched nearly every profession and discipline, growing exponentially in recent years and showing infinite potential when it comes to things as simple as internet searches or as complex as brain surgery. For officials with the Arizona Supreme Court, their venture into AI is rooted in a desire to promote trust and confidence in the judicial system.

What helped solidify the court’s need for more public outreach?

There was a protest outside the state Capitol last April and calls for two justices to be booted after the Arizona Supreme Court ruled that a Civil War–era law that banned nearly all abortions, except when a woman’s life is in jeopardy, could be enforced. Emotions flared on both sides of the issue.

When Chief Justice Ann Timmer took over the court last summer, she made public trust a key pillar of her platform. She had already been thinking about ways to reach out to the public using digital media for a few years, and the abortion ruling, among other rulings, helped her to solidify the idea that the court needs to be part of the narrative as people learn about opinions and what they mean.

“We serve the public better by saying, OK, we’ve issued this decision,” she said. “Now, let us help you understand what it is.”

Timmer told the Associated Press earlier this year that if the court had to do the abortion ruling over again, it would have approached the dissemination of information differently. In a Wednesday interview, she said that a news release and avatar video could have helped the public better understand the legal underpinnings of the lengthy decision—possibly including what it didn’t do, which she said some misunderstood.

“We got a lot of backlash for it and probably deservedly so, in terms of how can we complain that people don’t understand what we did when we didn’t really do enough to give a simplified version,” she said in the January interview, explaining that people want to know the basis for the court’s decisions and what they can do, such as lobbying state lawmakers for whatever changes in law would support their positions.

Democratic Gov. Katie Hobbs signed a repeal of the ban last May, and in November, Arizona voters approved a constitutional amendment expanding abortion access up to the point of fetal viability.

Who are Daniel and Victoria, and how do they work?

Created with a program called Creatify, Daniel and Victoria in a way bring to life the court’s news releases. Videos featuring one or the other are being posted for every ruling by the high court, and may be used for Access to Justice projects, community programs and civics information in the future.

The court has been sending out releases since October to summarize and explain rulings. After seeing success with the releases, it began exploring options to convey that information through video.

The AI-generated avatars were the most efficient way to produce videos and get the information out, said court spokesperson Alberto Rodriguez. Producing a video usually can take hours, he said, but an AI-generated video is ready in about 30 minutes. The court might introduce more AI-generated reporters in the future, Rodriguez said in a news release.

The justice who authors the legal opinion also drafts a news release, the wording of which must be approved by the entire bench. The justice then works with the court’s communications team to craft a script for the avatars—the avatars aren’t interpreting original court decisions or opinions, Rodriguez said.

Daniel and Victoria’s names and physical appearances were designed to represent a wide cross-section of people, Rodriguez said. He said they aren’t meant to come off as real people and the court emphasizes their AI origins with disclaimers. The court is exploring different emotional deliveries, cadences, and pronunciations as well as Spanish translations for the avatars, Rodriguez said.

Will the avatars resonate with their audience?

Mason Kortz, a clinical instructor at the Harvard Law School’s Cyberlaw Clinic at the Berkman Klein Center for Internet & Society, described the court’s new cyber employees as “quite realistic.” While their voices might give them away, he said some people could be fooled into thinking that Daniel and Victoria are real reporters if viewers are only reading the subtitles and looking at the characters’ movements and facial expressions.

Kortz also said it would be better for the language of the disclaimer that is in the videos’ text description to be featured more prominently.

“You want to make it as hard as possible for someone to advertently or inadvertently remove the disclaimer,” he said.

Asheley Landrum, associate professor at the Walter Cronkite School of Journalism and Mass Communication at Arizona State University, said the avatars feel robotic. She said a format that mimics real dialogue and storytelling might be more engaging than an AI reading of a news release.

“Because it’s not just about using AI or even creating videos,” she said, “but about doing so in a way that really resonates with audiences.”

Still, it’s fine line. She said engaging characteristics can help to build trust over time but the danger is that content could appear biased.

—Sejal Govindarao, Associated Press

View the full article