When Google fired one of its leading artificial-intelligence computer scientists recently, the media blowback was sharp and extensive–and raised several complex questions for HR leaders.
Timnit Gebru, an expert researcher in how facial recognition applications can be biased against people of color, was recently terminated after collaborating with Google colleagues on an academic paper exploring the potential for bias in AI systems, according to the New York Times. Prior to her termination, Google had asked that Gebru withdraw her authorship (and that of three other Google researchers) from an academic paper on the issue of AI and discrimination. In the same Times article, she said she told leadership she “would resign after an appropriate amount of time if the company could not explain why it wanted her to retract the paper and answer other concerns.” Google then told her it accepted her resignation immediately, according to the article.
Gebru, who served as co-leader of Google’s Ethical Artificial Intelligence Team, told Bloomberg that Google’s actions represented “the most fundamental silencing.” She has also publicly speculated that a group email in which she criticized the company’s diversity programs may have played a role in her firing.
Google officials did not respond to requests for comment, according to those media outlets. Sundar Pichai, CEO of Google parent company Alphabet, has since sent an email to employees apologizing for the fallout of the termination and promising an investigation, but not going so far as saying the company erred in the firing.
While the facts of the situation have yet to be sorted out and there has been no litigation mentioned, AI expert John Sumser has a strong view of the situation reported so far.
“Our journey to a set of AI tools and processes that work without penalizing certain classes of people is a long road,” says Sumser, founder, principal author and editor-in-chief of HRExaminer and a columnist for HRE. “Unfortunately, Google’s treatment of Timnet Gebru tells the world that focusing on the actual problem will be dealt with in a severe way.”
In this case, “someone who was paid to notice the emperor’s clothing noticed out loud that the emperor was naked. So, the emperor fired her,” Sumser says.
If real progress is to be made toward bias-free technology or the advancement of human rights, decision-makers like Google must embrace transparency, even if it’s embarrassing, he says.
“I hope that HR tech vendors will take the time to tell their clients and the market that they will not engage in this sort of ‘marketing is more important than substance’ kind of behavior,” he says.
Beyond the repercussions for HR technology, the Google situation also raised a number of legal questions. Ayesha Krishnan Hamilton, an employment lawyer in Princeton, N.J., cautions that, when facing a termination decision, employers must assess exposure that may likely arise from the termination.
For instance, she explains, employers must determine whether the employee fits into a protected classification (race, gender, age, disability, etc.) or whether that employee has engaged in some protected activity (i.e. as a whistleblower).
“The employer must also document and prove the legitimate business reasons that justified the termination to show that it was not motivated by retaliation or discrimination,” she says, adding that, in particular, the company’s investigator should pay close attention to the timing of events leading up to the termination.
In the Google case, she says, the facts as reported in the media may not rise to the level of having a winning retaliation and/or racial bias claim. Right now, there’s not enough information to determine whether the fired AI researcher was publicly complaining about any violation of law or public policy by Google, which would give rise to whistleblower protection, Hamilton says.
“The facts certainly do not tend to indicate that she was fired solely because of her racial background,” she adds. “While the facts of her case may be upsetting and bad PR for Google, they may not rise to the level of conferring liability upon the company.”
“Companies need to learn how to do both,” she says.
Also, they need to understand that the two “d-words,” discrimination and diversity, are not interchangeable.
“Discrimination is illegal; a lack of diversity is not,” she says.