Search About Newsletters Donate
Support independent, nonprofit journalism.

Become a member of The Marshall Project during our year-end member drive. Our journalism has tremendous power to drive change, but we can’t do it without your support.

A new study in the journal Language found that witnesses who speak in cultural or regional dialects are more likely to be misunderstood in the courts.
Justice Lab

In Court, Where Are Siri and Alexa?

When it comes to setting the record straight, court reporting technology is still not up to speed.

Court reporters turn out millions of transcripts a year documenting what’s said in legal proceedings. But in the high-tech age of Siri and Alexa, why aren’t courts turning to technology to produce the records in real time?

That’s one of the questions that has been raised after a new study found court reporters in Philadelphia couldn’t accurately transcribe the speech of some African Americans, instead producing records that were “ungrammatical and nonsensical.” The study, set to be published in the journal Language, found that witnesses who speak in cultural or regional dialects are more likely to be misunderstood in the courts.

That’s a big problem in a system that requires transcripts for appeals of court decisions. And what’s heard—or, in this case, incorrectly heard—in the courtroom can affect the outcome of a case.

Fourteen states use audio or video recorders in lieu of court reporters, according to a 2015 study by the National Center for State Courts. Many of the states cite cost as the reason for the shift to digital. In 2009, the Iowa Judicial Council conducted a sample study of a dozen courts across five states and found that digitally recorded trials would save the Iowa courts over $10 million each year in court reporter salaries. In Utah, which did away with court reporters in 2009, the state saved $1.3 million a year by eliminating court reporters almost entirely, according to a 2012 study conducted by the NCSC and the State Justice Institute.

But technology has not completely eliminated humans from the process of making court transcripts. In courts that digitally record trial proceedings, humans are still tasked with turning the tape into readable text documents.

The Iowa Council’s study in 2009 also reviewed the possibility of using a technology that could transcribe in real time, often called speech to text dictation, and concluded that the technology was “not sufficiently advanced.” Since then, speech to text software—the kind used by Siri or Alexa when you dictate a text message, for example—has dramatically improved. But researchers caution the technology still isn’t accurate enough, especially when dealing with dialects and accents.

Alexa and other similar programs leverage the internet to learn commonly asked commands, which are often simple, like the directions one might give to a foreign traveler looking for the nearest train station. But conversational speech is much more dynamic, making it difficult for dictation tech to keep up.

“When you talk to Alexa you are mostly using the same five sentences. Turn the light off, or order me this. Play this song,” said Gerald Friedland, an adjunct associate professor in electrical engineering and computer science at the University of California, Berkeley. “The moment you go from humans talking to computers to humans talking to humans, things get much harder.”

To ensure accuracy in a courtroom setting, Friedland said, voice recognition technology would have to be trained on commonly used words and phrases. In time, as the technology continues to develop, Friedland suggests transcription tech could act as an aid to court reporters.

Audio tapes of trials already serve as a beneficial redundancy, according to proponents of tech in the courtroom. In the event the written transcription is contested, or a portion is missing, lawyers can consult the tapes and try to piece together what was said. In contrast, if a court reporter gets something wrong, or leaves out testimony, there is no back-up.

In Connecticut, which uses audio recordings, transcript revisions are rare. Only 30 of the 17,000 transcripts requested last year were challenged for a discrepancy between the written record and the audio recording. In these cases, court officials tested the transcript against the audio recordings, and made revisions in 13 cases, according to Melissa Farley, a court spokesperson.

The key question is not whether technology can outperform court reporters, but rather, what will result in the fairest trial for defendants, said Drew Findling, president of the National Association of Criminal Defense Lawyers.

When it comes to real-time disputes about what was said, having a court reporter to read back notes is critical, Findling said. And in complex cases or cases with multiple defendants, getting a daily transcript is crucial for keeping track of each defendant’s testimony, cross-examining witnesses or for planning closing arguments.

“When you have a transcript, you don’t have to have the jury relying on their collective memory of the testimony. If you get a piece of the transcript, you can put it on a Powerpoint, and say, ‘You may remember one week ago when John Doe said this,'” Findling said. “There will be no one popping up, saying, ‘I object, that’s not what he said.’”

Whether audio tapes have improved court records is debatable, said Roan Anne, a private criminal defense lawyer in Colorado. Anne said she has seen transcripts made from audio recordings with chunks missing because something was inaudible or garbled. She cautions that incoherent and incomplete transcripts make it harder to craft appeals, resulting in a breakdown of trust in the justice system.

“The system is designed to ensure and guarantee that people convicted of crimes were convicted in a fair proceeding,” she said. “If the system is organized in such a way that meaningful appellate review is somehow impossible, the justice system suffers, and [so does] the public’s confidence” in the system.

Some argue having a court reporter can ensure that everyone is heard, as transcripts made from audio can be lost if there is inaudible talk or if the recording device wasn’t turned on. In those instances, portions of the trial are lost forever.

“Lawyers tend to forget that when in court they are making a record, and get lost in the heat of the moment,” said Nate Alder, an attorney with Christensen and Jensen in Salt Lake City, Utah. “If you were to go to a deposition, court reporters are always reminding lawyers to speak one at a time. I think the highest level of accuracy is a court reporter slowing everyone down and making sure everyone takes their time.”