By Lauren Weber and Deepa Seetharaman
By her second day on the job, Sarah Katz knew how jarring it can be to work as a content moderator for Facebook Inc. She says she saw anti-Semitic speech, bestiality photos and video of what seemed to be a girl and boy told by an adult off-screen to have sexual contact with each other.
Ms. Katz, 27 years old, says she reviewed as many as 8,000 posts a day, with little training on how to handle the distress, though she had to sign a waiver warning her about what she would encounter. Coping mechanisms among content moderators included a dark sense of humor and swiveling around in their chairs to commiserate after a particularly disturbing post.
She worked at Facebook's headquarters campus in Menlo Park, Calif., and ate for free in company cafeterias. But she wasn't a Facebook employee. Ms. Katz was hired by a staffing company that works for another company that in turn provides thousands of outside workers to the social network.
Facebook employees managed the contractors, held meetings and set policies. The outsiders did the "dirty, busy work," says Ms. Katz, who earned $24 an hour. She left in October 2016 and now is employed as an information-security analyst by business-software firm ServiceNow Inc.
Deciding what does and doesn't belong online is one of the fastest-growing jobs in the technology world -- and perhaps the most grueling. The equivalent of 65 years of video are uploaded to YouTube each day. Facebook receives more than a million user reports of potentially objectionable content a day.
Humans, still, are the first line of defense. Facebook, YouTube and other companies are racing to develop algorithms and artificial-intelligence tools, but much of that technology is years away from replacing people, says Eric Gilbert, a computer scientist at the University of Michigan.
Facebook will have 7,500 content reviewers by the end of December, up from 4,500, and it plans to double the number of employees and contractors who handle safety and security issues to 20,000 by the end of 2018.
"I am dead serious about this," Chief Executive Mark Zuckerberg said last month. Facebook is under pressure to improve its defenses after failing during last year's presidential campaign to detect that Russian operatives used its platform to influence the outcome. Russia has denied any interference in the election.
Susan Wojcicki, YouTube's chief executive, said this month that parent Google, part of Alphabet Inc., will expand its content-review team to more than 10,000 in a response to anger about videos on the site, including some that seemed to endanger children.
No one knows how many people work as content moderators, but the number likely totals "tens of thousands of people, easily," says Sarah Roberts, a professor of information studies at the University of California, Los Angeles, who studies content moderation.
Outsourcing firms such as Accenture PLC, PRO Unlimited Inc. and Squadrun supply or manage many of the workers, who are scattered among the corporate headquarters of clients, work from home or are based at cubicle farms and call centers in India and the Philippines.
The arrangement helps technology giants keep their full-time, in-house staffing lean and flexible enough to adapt to new ideas or changes in demand. Outsourcing firms also are considered highly adept at providing large numbers of contractors on short notice.
Facebook decided years ago to rely on contract workers to enforce its policies. Executives considered the work to be relatively low-skilled compared with, say, the work performed by Facebook engineers, who typically hold computer-science degrees and earn six-figure salaries, plus stock options and benefits.
Reports of offensive posts from users who see them online go into a queue for review by moderators. The most serious categories, including terrorism, are handled first.
Several former content moderators at Facebook say they often had just a few seconds to decide if something violated the company's terms of service. A company spokeswoman says reviewers don't face specific time limits.
Pay rates for content moderators in the Bay Area range from $13 to $28 an hour, say people who have held jobs there recently. Benefits vary widely from firm to firm. Turnover is high, with most content moderators working anywhere from a few months to about a year before they quit or their assignments ended, according to interviews with more than two dozen people who have worked such jobs.
Facebook requires that its content moderators be offered counseling through PRO Unlimited, which actually employs many of those workers. They can have as many as three face-to-face counseling sessions a year arranged by an employee-assistance program, according to an internal document reviewed by The Wall Street Journal.
The well-being of content moderators "is something we talk about with our team members and with our outsourcing vendors to make clear it's not just contractual. It's really important to us," says Mark Handel, a user research manager at Facebook who helps oversee content moderation. "Is it enough? I don't know. But it's only getting more important and more critical."
Former content moderators recall having to view images of war victims who had been gutted or drowned and child soldiers engaged in killings. One former Facebook moderator reviewed a video of a cat being thrown into a microwave.
Workers sometimes quit on their first or second day. Some leave for lunch and never come back. Others remain unsettled by the work -- and what they saw as a lack of emotional support or appreciation -- long after they quit.
Shaka Tafari worked as a contractor at messaging app Whisper in 2016 soon after it began testing a messaging feature designed for high-school students. Mr. Tafari, 30, was alarmed by the number of rape references in text messages he reviewed, he says, and sometimes saw graphic photos of bestiality or people killing dogs.
"I was watching the content of deranged psychos in the woods somewhere who don't have a conscience for the texture or feel of human connection, " he says.
He rarely had time to process what he was seeing because managers remotely monitored the productivity of moderators. If the managers noticed a few minutes of inactivity, they would ping him on workplace messaging tool Slack to ask why he wasn't working, says Mr. Tafari.
A Whisper spokesman says moderators were expected to communicate with managers about breaks. Whisper no longer employs U.S.-based moderators. It uses a team in the Philippines along with machine-learning technology.
A former content moderator at Google says he became desensitized to pornography after reviewing content for pornographic images all day long. "The first disturbing thing was just burnout, like I've literally been staring at porn all day and I can't see a human body as anything except a possible [terms of service] violation," he says.
He says the content moderators he worked with were hit hardest by images of child sexual abuse. "The worst part is knowing some of this happened to real people," he says.
Facebook says it removes abusive content, disables accounts and reports some types of suspected criminal activity to law-enforcement agencies, the National Center for Missing and Exploited Children and other groups.
A spokeswoman for Facebook says it offers counseling, resiliency training and other forms of psychological support to everyone who reviews content.
A YouTube spokeswoman says: "We strive to work with vendors that have a strong track record of good working conditions, and we offer wellness resources to people who may come across upsetting content during the course of their work."
Two Microsoft Corp. employees who reviewed content allege the company failed to provide a safe workplace and support after it became clear that viewing sexual abuse and other material was traumatizing them. They say they regularly reviewed material involving the sexual exploitation of children and adults.
In a civil lawsuit filed in a state court in Washington, the employees said they suffered from insomnia, anxiety and debilitating post-traumatic stress disorder. One of the Microsoft managers, Henry Soto, said he has found it difficult at times to be near computers or his own son.
The two employees are on medical leave from Microsoft. They are seeking damages for past and future emotional distress, mental-health treatment, impaired earning capacity and lost wages. A trial in the case has been scheduled for October.
Rebecca Roe, a lawyer for the other manager, Greg Blauert, says tech companies should be held accountable for the well-being of content moderators. Contractors are especially at risk because they have little job security and are less likely to seek help, she says.
Microsoft denies it failed to provide a safe workplace. The company says it "takes seriously its responsibility to remove and report imagery of child sexual exploitation and abuse being shared on its services, as well as the health and resiliency of the employees who do this important work." Content moderators at Microsoft are required to participate in psychological counseling.
While Facebook and other tech companies are spending more money than ever to police their sites, few seem inclined to convert outside contractors into employees. Some content moderators say they have been told by managers that data based on their decisions would help train algorithms that might eventually make the humans expendable.
Andrew Giese, 28, says he left his full-time job at grocery chain Trader Joe's Co. to do content moderation for Apple Inc., where he spent a year mostly sifting through user comments posted on iTunes as a contract worker.
When the contract ended, he says, Apple managers told him he would be eligible to return after a three-month furlough, but there were no openings available. Mr. Giese eventually went back to Trader Joe's.
(MORE TO FOLLOW) Dow Jones Newswires