VentureBeat presents: AI Unleashed - An exclusive executive event for enterprise data leaders. Network and learn with industry peers. Learn More
University of Washington language researchers and legal professionals recently created a labeled dataset for detection of interruptions and competitive turn-taking in U.S. Supreme Court oral arguments. They then used the corpus of “turn changes” to train AI models to experiment with ways to automatically classify turn changes as competitive or cooperative as a way to analyze gender bias.
“In-depth studies of gender bias and inequality are critical to the oversight of an institution as influential as the Supreme Court,” reads the paper University of Washington researchers Haley Lepp and Gina-Anne Levow published on preprint repository arXiv one week ago. “We find that as the first person in an exchange, female speakers and attorneys are spoken to more competitively than are male speakers and justices. We also find that female speakers and attorneys speak more cooperatively as the second person in an exchange than do male speakers and justices.”
Attorneys who speak before the Supreme Court are allotted 30 minutes of oral argument and are expected to stop talking when a justice speaks. Linguists have observed men interrupting women routinely in professional environments and other settings.
Turn changes are defined as instances when one person stops speaking and another person starts speaking. Short audio clips of each turn change were annotated as competitive or cooperative by 77 members of the U.S. legal community who identify as an attorney, judge, legal scholar, or law student in their second year or higher. Lepp and Levow’s work focuses on measuring whether the turn change was cooperative or competitive, based on oral argument audio the Supreme Court made available, in part because previous work by Deborah Tannen found that interruptions in speech can be part of regular discourse and that the context of the conversation can be a factor.
An exclusive invite-only evening of insights and networking, designed for senior enterprise executives overseeing data stacks and strategies.
The paper devoted to gender bias analysis was published days before the death of Supreme Court Justice Ruth Bader Ginsburg at the age of 87. Ginsburg was the second woman ever appointed to the U.S. Supreme Court. As a litigator for the American Civil Liberties Union (ACLU), Ginsburg successfully argued cases before the Supreme Court that greatly extended women’s rights in the United States. On Friday, she will be the first woman and the first Jewish person in U.S. history to lie in state at the U.S. Capital building for members of the public to say goodbye. She was the longest-serving female justice in U.S. history.
Although voting has already begun in some parts of the country and Ginsburg pleaded in her final days to let the winner of the presidential election fill her vacancy, President Trump is expected to nominate a pick to fill her seat Friday or Saturday. Two Republican Senators pledged not to vote until the presidential election is decided, but Senate Majority Leader Mitch McConnell said just hours after her death that the president’s nominee will get a vote.
Details of the turn changes corpus dataset follow a 2017 study that used automation to identify the number of interruptions that occurred from 2004-2015. The study “Justice, Interrupted: The Effect of Gender, Ideology and Seniority at Supreme Court Oral Arguments” by Tonja Jacobi and Dylan Schweers found that women are interrupted three times as often as male Supreme Court justices are. Female Supreme Court justices were interrupted by attorneys as well as other Supreme Court justices, led by Anthony Kennedy, Antonin Scalia, and William Rehnquist. Scalia and Stephen Breyer also interrupted each other a lot.
A producer of the podcast More Perfect noticed people repeatedly interrupting Ginsburg, which led to an episode on the subject. Jacobi spoke on the podcast and said Ginsburg developed tactics to adapt to frequent interruptions, first by asking to ask a question, then pivoting to ask questions more like male justices who interrupt.
The episode also highlighted that Justice Sonia Sotomayor was found to speak as often as men in the Jacobi study, but has still drawn criticism from media commentators at times for being aggressive. Gender is pervasive in coverage of Supreme Courts, according to a 2016 analysis of media coverage in five democratic countries. The analysis found that generally women who ask questions like male justices are labeled abrasive, militant, or mean by critics.
Last year, the U.S. Supreme Court introduced a rule that justices will try to give attorneys two minutes to speak without interruption at the start of oral arguments.
VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.