Google Researcher Accused of Sexual Harassment, Roiling AI Field

(Bloomberg) -- Sexual harassment accusations have hit another corner of the tech industry, with allegations involving prominent artificial intelligence researchers, including one at Google, a leader in the field.

Data scientist Kristian Lum wrote in a blog this week that a man she called "S" grabbed her inappropriately at an industry conference in 2010 and said he took advantage of another woman she didn’t identify on separate occasions. Two people who were told about the conduct from two alleged victims told Bloomberg the man is Steven Scott, a senior researcher at Google.

Lum also wrote that a well-respected academic touched her inappropriately on the leg at the same conference and later sent her innuendo-laced messages. The man was later identified as Bradley Carlin, an expert in biostatistics, by people familiar with the alleged conduct.

The allegations were widely discussed on social media, drawing supportive comments and similar stories from researchers in academia and leading tech companies, including Alphabet Inc.’s Google and Microsoft Corp. Eric Horvitz, a Microsoft research director, called Lum’s post a "powerful and important report."

"I’d love to tell my side of the story, but I’m afraid you’ll need to get it from press@google.com," Scott wrote in an email, referring to Google’s public relations department. 

"Google is investigating the matter," Gina Scigliano, a company spokeswoman, said.

Scott has been suspended from Google as of Friday, according to a person familiar with the matter.

Carlin, a professor at the University of Minnesota, referred questions to the school, which said it is aware of the accusations but declined to comment further. 

The researchers involved are experts in Bayesian statistics, which underpins a powerful type of AI known as machine learning. The accusations have surfaced during a growing debate over the lack of diversity among machine learning researchers and whether computer scientists are paying enough attention to bias – including gender and racial bias – in the data sets they are using to train AI systems.

The allegations came to light after the recent conference on Neural Information Processing Systems (NIPS) in Long Beach, California. Some attendees complained that aspects of some parties held during the conference were inappropriate and potentially offensive to women. They also condemned a joke about sexual harassment made by Carlin, a member of an amateur band that performed at the NIPS closing night party.

Carlin, the band’s keyboardist, joked about the sexual misconduct allegations that forced the resignation of U.S. Senator Al Franken. After several attendees complained on Twitter that making light of sexual harassment was inappropriate, Carlin apologized on the band’s Facebook page.

Lum, who was not at the conference, learned of the incident and then wrote in the blog that she was “unsurprised to learn that a person involved in making the troubling comments is a well-respected academic who is widely known to behave inappropriately at conferences.”

Without naming Carlin directly, Lum said the academic had touched her inappropriately on the leg during an informal presentation she gave at a Bayesian mathematics conference in 2010. He also commented on her dress, saying it “was too sexy” for such an academic talk. Over the ensuing years, she wrote, the person sent her several innuendo-laced or inappropriate private Facebook messages, including messages in which he discussed watching pornography.

"I would like to defend myself against her accusations, but the matter has already been referred to my university’s EEO team, which I’m told will be conducting a full investigation into the matter," Carlin wrote in an email to Bloomberg, referring to the University of Minnesota’s Equal Employment Opportunity department. "So I’ve been instructed not to say anything more publicly pending the results of that investigation."

Evan Lapiska, a spokesman for the university, said the school is "aware of public accusations" involving Carlin but declined to comment further citing privacy protections. 

Lum declined to comment. In her post, she detailed further incidents of alleged harassment at the same 2010 conference. The "worst offender," she wrote, was a researcher that she only identifies as "S." At the end of the conference, while she was swimming with other researchers, the person grabbed Lum without her consent and put “his hands on my torso, hips, and thighs,” according to the post. Other female mathematicians spoke similarly to Lum about S, Lum added. And she wrote that he had "taken advantage of a junior person" under the influence of alcohol.

She said that several years later at another conference, during a reception hosted by Google, she overheard the same person bragging to friends about “banging smokin’ hot chicks.” When Lum gave him a castigating stare, he told Lum she was only jealous he hadn’t been talking about her, according to the blog.

"I knew it would be even more difficult to get people to find S’s behavior problematic since he is employed by a large tech company," she wrote.

The person in Lum’s post is Scott, a director of statistics research at Google, according to two people familiar with the situation. Katherine Heller, an assistant professor at Duke University, recognized Scott in Lum’s description and told Bloomberg that he had acted inappropriately with a former student of hers. Heller also said that several other female researchers had reached out to her with similar stories about other men in the field after Lum’s blog post was published. 

"There really is just a lot of sexual harassment of women in Bayesian statistics and machine learning," Heller said. Another person in the field said she had witnessed Scott’s actions with women in the past, but asked not to be identified given the sensitivity of the topic.

“Thank you for having the courage to speak out about this terrible behavior," Jeff Dean, head of Google’s Brain AI unit, tweeted to Lum. "This has no place in science, math, statistics, computing, or anywhere else.” Dean was commenting generally and he didn’t respond to an email asking about Scott.

Lum said that after she saw in October that "S" had been nominated as a candidate for the board of the International Society for Bayesian Analysis (ISBA), she brought concerns about his behavior to the organization’s president and he was removed from the ballot.

Scott, who has been at Google since 2008, was a candidate for the board and is no longer on the ballot, according to ISBA’s website and one person with knowledge of the organization who spoke to Bloomberg.

Kerrie Mengersen, ISBA’s president, did not respond to questions about Scott. But she said in an email that she and ISBA’s executive committee supported Lum coming forward with her account. “The Executive Committee takes any report of inappropriate behavior very seriously,” she said. “We have a great Society and a proud record of including women as members and as leaders. However, like many societies, we have recognized that we need to do more.”

Mengersen said the organization was in the process of establishing protocols for appropriate behavior and support mechanisms for members who have been harassed.

To contact the reporters on this story: Mark Bergen in San Francisco at mbergen10@bloomberg.net, Jeremy Kahn in London at jkahn21@bloomberg.net.

To contact the editors responsible for this story: Tom Giles at tgiles5@bloomberg.net, Alistair Barr, Molly Schuetz

©2017 Bloomberg L.P.