Bangalore: The world’s largest election has become something of a test case in how technology giants handle fake news after years of scandal. It’s not working out so well.
India has as many as 900 million voters in an election that culminates this week, with Prime Minister Narendra Modi’s ruling coalition headed for apparent victory. Particularly challenging for would-be fact-checkers, from Facebook Inc. to Google, is the country’s 23 official languages. Facebook has hired contractors to verify content in 10 of those languages, but those staffers are spread thin and posts in more than a dozen other languages — Sindhi, Odia and Kannada among them — are completely unvetted.
Pallavi Mishra has seen the scale of the challenge first-hand. The manager at Vishvas News, Facebook’s largest Indian-language fact-checking contractor, spent two weeks recently talking with internet users in small cities. She found most people are so new to social media they have no clue about bogus content. They share stories indiscriminately, with stupefying speed. “Being the ‘first’ to share things in their circles gave them a rush,” she says.
Facebook, Twitter Inc. and Google parent Alphabet Inc. are discovering the harsh reality that disinformation and hate speech are even more challenging in emerging markets than in places like the U.S. or Europe. A new category of users, recently digital, believe almost whatever they receive — especially if it comes from family or friends. Hundreds of millions read in languages the American tech giants haven’t even begun to monitor.
“Disinformation is spreading like wildfire in these parallel digital universes,” said Bharat Gupta, chief executive officer of Jagran New Media, which runs Vishvas News. “It’s a dark space that nobody talks about.”
About 90% of internet users coming online today are non-English speakers, often with more trusting attitudes than experienced surfers. In one Hindi example, a phony Facebook post for job vacancies at the subway train company in the city of Jaipur asked those interested to send personal details. Within hours, about 20,000 had responded with names, emails and phone numbers.
Efforts to monitor native languages have gotten off to a slow start. Social networks, messaging apps and content aggregators have started drafting community guidelines, automated tools are being built, and in-house content moderation teams created. Facebook is the first to make a commercial commitment, getting a jump on foreign rivals. Still, Vishvas began fact-checking Hindi Facebook posts just last December, while contractors in Punjabi and Urdu barely began a month ago.
Local-language speakers aren’t waiting to get online of course. Hundreds of millions who live in remote Indian villages are now zooming along cyber highways thanks to cheap smartphones and rock-bottom wireless prices. Native language internet users are forecast to make up three quarters of the country’s online users by 2021. And Hindi Internet users alone will outnumber English Internet users by that year.
Vishvas’ office is in a glass fronted low-rise on a narrow street in a New Delhi suburb called Okhla. The fact-checking outfit is part of Jagran Prakashan Ltd., a publicly traded media company that publishes India’s most-read newspaper, the Hindi language Dainik Jagran. Vishvas’ operation occupies one corner of a bay on the second floor. Orange partitions divide the workspaces, and vivid balloons, cricket paraphernalia and other personal adornments offer a pop of color against the pale vastness of a space filled with digital content creators.
Vishvas has five Hindi fake-news busters, and one each in Punjabi and Urdu. It’s a puny shield against the swamp of political misinformation, fraud and pornography. In this year’s election, malicious political propaganda often comes from shadowy cyber-armies targeting the ruling Bharatiya Janata Party, its main opponent Congress Party or the dozens of regional parties.
Hundreds of texts, audio and video clips are flagged daily to Vishvas’ fact-checkers by Facebook’s automated tool, or through dedicated email or WhatsApp helplines. At 9:30 each morning, the team chooses two dozen stories to debunk from among hundreds mostly related to the elections, political parties or prominent leaders. Verifications begin with reaching out to the subject or their PR representative. A variety of tools like image recognition and sun-shadow calculators are employed. By the day’s end, about 20 pieces of fake content are debunked in detailed articles on Vishvas’ website. Facebook drastically reduces their distribution and they are fed to its machine-learning algorithms.
Bhagwant Singh, who joined Vishvas last month to bust fakes in Punjabi, is still learning the ropes. The 22-year-old, who sports a turban and beard in keeping with his Sikh religious faith, checked out a photo of Rahul Gandhi that had gone viral online. It showed the president of the Indian National Congress with a badly-bruised face, along with the claim he had been brutally attacked by voters in his parliamentary constituency Amethi where he is locked in a fierce battle with a minister from the BJP. The truth? A portion of a photo of an unknown man’s injured face had been skillfully photo-shopped onto Gandhi’s image.
Singh also debunked a post with an image of Bollywood actor Sunny Deol, a BJP contestant, standing next to the chief of the right-wing RSS, a BJP affiliate unpopular in Punjab. “Par phikar na karo Gurdaspur,” Singh read out from the Facebook post. “Ye chhethi hi laha denge” (Don’t worry Gurdaspur, the people will make him fall). “Social media is very polluted,” he says.
Sitting on the other side of the partition is Umam Noor, who says she is India’s first Urdu fact-checker. The baby-faced 23-year-old laughs at some of the frauds she has busted. A photograph of a random man at a street side umbrella repair shop had been tagged as the brother of former President A.P.J. Abdul Kalam. Despite his brother reaching the highest office, the family still held on to its humble profession, said the text. Noor’s investigation showed that President Kalam’s brother is a 102- year-old who leads a quiet life in his village. The umbrella repair man was not remotely connected.
“Nobody checks,” says Noor. “They want to show off by just forwarding everything they receive.” This particular day she also vetted a photo claiming to be BJP men insulting Rohingya Muslim refugees in Kolkata city. The photo turned out to be of men from the other side of the country protesting India’s new taxation system. Noor’s family thinks her work is not just trendy but vital. “Because of my job, I’ve gained new respect in my family and friends’ circles,” she says.
Whatever the result in India’s election, there’s no denying that fake news has amplified rage and polarization. The problem will outlast the election and worsen in the coming months, said Rajesh Upadhyay, editor-in-chief of Vishvas News.
“Those riding a bullock cart find themselves overnight behind the wheel of an automatic car,” he says. “How can the problem go away?”
Also read: How Facebook is fighting fake news in the Lok Sabha elections