Welcome to the front line of the global battle against fake news.
Agence France Presse is a French government-owned international news service. Its New Delhi bureau fact-checking operation is just one of dozens that Facebook has started paying for around the world this year. It’s a fight Mark Zuckerberg and his executive team have come to reluctantly and, some would argue, too late. But with the $US400 billion company’s reputation being battered by growing worldwide concern over manipulation of the truth, it is amassing the troops.
For Malakunas, who has worked in Asia for 20 years and is a former Manila bureau chief, the work has proved a revelation. Establishing operations to investigate dubious news stories, Photoshopped images and inflammatory online posts in Islamabad, Jakarta and Manila has made him look more deeply into what is appearing on social media.
“Ahead of the elections it’s so important to understand what’s happening underneath. It’s not just Facebook, but across all social media. This is where the rise of the populists is coming from, and the mainstream media hasn’t picked any of them.”
To illustrate the point he cites the account of an imposter impersonating a famous Muslim cleric in Indonesia. “The imposter account has 700,000 followers!”
Ultimately, he says, when people are going to the polls and casting their ballots based on false information, it is unequivocally a threat to democracy.
Over in the Philippines, another AFP fact-checker has called out a post declaring late Philippine dictator Ferdinand Marcos held a Guinness World Record for being the ‘Most Brilliant President in History’. It would be ridiculous except the post is on a page with nearly one million followers which supports Ronald dela Rosa, the Philippines’ former police chief who is running for a Senate seat in next year’s elections. Along with the Marcos family, dela Rosa is an ally of the country’s strongman president Rodrigo Duterte.
“Most journalists would snigger and laugh and say, ‘As if anyone would believe it’,” says Malakunas. “But then you see how much traction it gets and the way it can change how people are thinking.
“This is a very different way to get people to support a political party than what we are traditionally used to. In an election, we are still focused on what the prime minister and the opposition say. But there is an underworld of cyber-fakery and it is having a significant impact on society. It is shaping views to an extent that I don’t think any of us really know.”
Forced to change
For years, Zuckerberg played down Facebook’s responsibilities in managing content published on the site. People had a right to say things, he would often insist in interviews, even if they were wrong. His mission was to connect people, not to tell them what to share. His focus was on growing the business, which started in a Harvard dorm room almost 15 years ago. The way he did this is encapsulated in the company’s motto: “move fast and break things”. While Facebook had basic rules about nudity, hate speech and violence, Zuckerberg was reluctant to interpose a value system on it or be forced into an arbiter-of-truth role.
And it worked for Facebook if posts were controversial. The business model is built on traffic – the more users there are, the higher the advertising revenue. Sensational tales and eye-catching photos and videos draw people in and keep them engaged on the site for longer.
But as the number of Facebook’s monthly active users rocketed past 2 billion and the company was damaged by a series of scandals – including Russian interference in the 2016 US presidential election, revelations Cambridge Analytica harvested the personal data of millions of users without their consent and criticism over its failure to stop the spread of hate speech that fuelled the targeting of Rohingya Muslims in Myanmar – the management team was forced to change its approach.
In the third quarter of this year, Facebook took down or slapped an adult-warning on 30 million pieces of content, including roughly three million instances of both terrorism propaganda and hate speech. The company also disabled a staggering 754 million fake accounts – most within minutes of registration and many of them set up by robots to scam and spam users.
“Facebook grew really quickly and I don’t think they ever stopped to consider the implications of this growth so that now they are absolutely playing catch-up,” says Jennifer Beckett, a social media and communications expert at the University of Melbourne. “That’s all they can do at this point.”
The scale of the challenge is huge. As Beckett points out, every minute on Facebook 510,000 comments are posted and 136,000 photos uploaded.
“This is an immense amount of content that’s going up on one platform globally, in multiple languages, in multiple social contexts and multiple political contexts and a lot of it has to be moderated by people because it relies on context and machines don’t do context.”
Facebook is not alone. Other social media platforms, such as Google, are also being forced to face up to their responsibilities as the power of tech giants in public debates, consumer decision-making and elections becomes ever clearer.
Inside the war rooms
It’s a Monday afternoon in Facebook’s tech-hip decorated Sydney office and Monika Bickert is apologising for speaking “a mile a minute”. The former US federal criminal prosecutor, who now heads up Facebook’s global policy management team and is responsible for policing content on the site, is on a charm offensive. She has a lot to say.
Bickert is travelling the world to showcase Facebook’s efforts to “create a safe community” by pulling down images and videos of child abuse and terrorism propaganda and slowing down the distribution of false reports. In quick succession, she outlines the steps the company has taken this year: refining its content standards and making those public; improving the technology to find and remove photos; building up a fact-checking network by partnering with traditional news organisations like AFP and Associated Press; and setting up election “war rooms” to ensure the sort of interference that took place in the 2016 US election doesn’t happen again. That’s important as countries including Indonesia, Australia, India and the Philippines will all hold key elections next year.
Bickert is a Harvard Law graduate who prior to Facebook worked as a legal advisor in the US embassy in Bangkok with a focus on child exploitation and human trafficking. She says when she started at the company in 2012 there were only a handful of people writing content policies and a couple of hundred content reviewers. Those teams have grown to more than 100 and 15,000, respectively. All up, Bickert says there are now 30,000 people working on safety and security at Facebook. What she doesn’t specify is that much of this growth – an extra 10,000 people – has happened in just the last five months, a tumultuous period for the company, which has been battling to salvage its image as a responsible platform manager and stave off an onslaught of new regulation.
The new employees – who are in addition to the third-party fact-checkers at organisations like AFP – are content reviewers, experts in counter terrorism, child safety and domestic violence, engineers, ArtificiaI Intelligence specialists and academics.
“What we’re fundamentally trying to do with our standards is keep the community safe while also ensuring that people have the ability to express themselves and, for what it’s worth, I don’t think those things are necessarily at odds with one another,” says Bickert.
We are in the meeting room at Facebook’s Sydney headquarters, which strives to be unconventional with graffiti art, plant dividers and bike racks on the office floor. There is a Facebook Wall in reception where visitors are invited to write down ‘what’s on their mind’, a feature in all of its global offices.
Bickert is explaining what has changed in the content policies and why making them public is important.
“Before April, we said with graphic violence if you are sharing content you can do so to raise awareness or condemn violence but we will remove it if you are sharing it to celebrate or glorify violence.
“Now, you can see exactly what we mean by extremely violent videos – they include videos of burying bodies or charred bodies, videos of dismemberments; the nitty gritty of exactly how we interpret these things so that people know where the line is.”
Those sorts of decisions are made at a fortnightly meeting when around 20 top executives gather around a long table at the company’s headquarters in Menlo Park, California and hook up with offices around the world via video conference. At each meeting, they discuss on average two recommendations and another two “heads up”, which are issues a team flags and then puts together a working group to decide what to do.
Bickert holds court at these hour-long sessions, quizzing her team about definitional grey zones, whether Facebook policies line up or differ from government guidelines and how changes might be received by the public.
The challenging tone of these meetings illustrates just how hard Facebook is finding it to define where the ‘line’ actually is, and who should be drawing it. Staff question decisions and the processes by which they are made. These include challenging the backgrounds of experts providing input amid concern there is not enough diversity and too many “white men” influencing recommendations. Others worry about the tough calls that will need to be made based on changes to definitions of child abuse, terrorism propaganda or hate speech.
In Sydney, Bickert pulls up a complicated “enforcement” graphic on a screen at the front of the room – a snapshot of how Facebook is doing – and walks over to point out some statistics, reminding me to move my recorder to pick up what she is saying.
In the third quarter of this year, Facebook took action on more than 30 million pieces of content, which means removing it or adding an adult-only warning. The company also disabled 754 million fake accounts and took action on about three million pieces of both hate speech and terrorist propaganda.
“We contract with third-party intelligence providers who will let us know ISIS have put out a new propaganda video. We can use software to take a digital fingerprint of that video, what we call a hash, and then using that hash we can stop it from ever hitting Facebook,” says Bickert, stressing how far technology has come in the last few years to detect this type of content.
“With terrorism and child exploitation imagery, we’re over 99 per cent in terms of how much of that content we flag but it’s much harder for hate speech,” she says, noting the company picks up just over half and the rest is from user reports.
Tall and confident, with a broad smile and an impressive grasp of the numbers, Bickert is a capable envoy but she is on a tough assignment. While the Californian is keen to talk about everything Facebook is proactively doing to clean up its platform, users, journalists, academics and rivals are focused on the company’s travails.
‘Don’t poke the bear’
Just five days before AFR Weekend sat down with Bickert in November, The New York Times published an investigation into the use of Facebook in propaganda campaigns. The report claimed security experts at Facebook had known about Russian interference in the 2016 US presidential election but management delayed telling the board and the public.
Bickert was also mentioned in the article after being delegated a decision on a Donald Trump post in 2015, along with Facebook’s departing head of communications and public policy Elliot Schrage and its vice president public policy Joel Kaplan. Together, the three executives decided not to suspend Trump’s account or remove a post in which he called for a ban on Muslim immigrants after it was flagged by Zuckerberg, who is a Democrat. Kaplan, a former George W. Bush staffer, argued it would look like they were obstructing free speech and reportedly said: “Don’t poke the bear.”
Bickert explains the contentious decision this way. “We make clear in the context of immigration, if people are calling for exclusion, we try to create space for that discussion on the site because we know that immigration policy is something that is a regular part of political speech.”
It’s these sorts of calls for which Facebook finds itself increasingly in hot water. They are complicated and subjective and attract emotional responses from both the right and left of the political spectrum.
From early on, Facebook has positioned itself as a platform and not a publisher but increasingly it finds itself facing similar challenges to traditional media organisations. That’s why its not surprising that Facebook, and also Google, is looking to work with those organisations to help resolve its credibility problem by stemming the flow of false information on its platform.
“It’s interesting Facebook have come back to the mainstream media organisations for fact-checking and they are paying for it, which puts a value on the work we do and is an acknowledgement of its importance,” says AFP’s Malakunas.
“Right now it feels like putting a thumb in the dam wall but more important is the attitude change from big tech. They have been seen as the major problem but they are also the solution.”
Malakunas says the way it works is Facebook-affiliated fact-checkers have access to a queue of suspect news flagged by users or uncovered by an algorithm. They can refer to the queue throughout the day and follow-up worthwhile leads – although they are under no obligation to clear the queue – and mark the reports if they are false. Facebook will then dial down the distribution of that post and if readers are still accessing it, direct them to other related news reports.
AFP uses social media monitoring platforms such as CrowdTangle to find leads and sets up What’sApp channels so that people can report misinformation and disinformation more easily. Increasingly, AFP publishes the fact-checks on its news wire and they are well read.
But are all of these efforts by Facebook and its fact-checkers enough? Melbourne University’s Beckett is doubtful. She says the task is too big and the company has a questionable history in acting on reports of misuse. She points to Facebook’s slow response to hate speech directed at Rohingya Muslims in Myanmar, who suffered violent attacks.
“Facebook was warned repeatedly by activists and academics what was happening in Myanmar,” she says. “It wasn’t until the UN stepped in and called the company out that they started to respond. That was either amazing naivety, pure stupidity or just ‘we don’t care’.”
Beckett says, like many people, she was a big believer in social media when it first took off only just over a decade ago.
It was going to solve the world’s problems, give a voice to those without one and connect people around the globe. But she has become more and more sceptical as platforms have been used to spread propaganda. Even Bickert concedes there is a problem.
“We know misinformation has been out there forever but what’s different about social media is that it’s a unique amplification tool,” she says. “We are trying to counter that virality and then provide information so that people can make well-informed decisions about what to trust.”
That’s where fact-checkers like AFP come in. “We are starting to work with third-party fact-checking organisations,” says Bickert. “This is something we’re building out. It’s not a complete program.”
Not all fact-checkers are happy with Facebook’s approach and there have been reports of some smaller partners pulling out. But AFP is backing the initiative and Malakunas says staff are reasonably upbeat about the impact it is having.
Facebook is currently looking into hiring a third-party fact checker in Australia but does not say whether that arrangement will be in place ahead of the election, which will be held before June next year. Bickert says the biggest issue that comes up in Australia is teenage bullying and Facebook has responded by training digital ambassadors and giving $1 million to Project Rockit, a local group which sends young people out to schools to encourage students to challenge cyberbullying.
Russell Skelton, who heads a local fact-check unit run jointly by RMIT University and the ABC, says Australia has fewer fake news websites than other countries because the aim of many sensational reports is to attract people to sites with heavy advertising.
“A lot of the fake sites are about generating money. We just don’t have enough people to make it financially viable,” he says.
“In countries where there is a strong national broadcaster like the BBC, ABC or Canadian Broadcasting Corporation, there tends to be less of this as people have a reference point to check things.”
Skelton says his fact-check unit’s remit is to scrutinise statements made by public figures trying to influence the public debate.
“When they tell porkies, we target those as a way of improving the discussion around important public policy issues. Then there is fake news, which is like a giant firehose with bots operating on an industrial scale spreading misinformation to millions of people in a matter of minutes. How do you counter that?
“I think Facebook and Google absolutely have a responsibility to identify and ping fake news. We call it misinformation and disinformation because fake news is a term that has been so abused by Trump that it is meaningless now.”
Adrian Turner, chief executive of the federal government’s data science agency, Data 61, says it’s “an open question whether these companies can get on top of it”.
He notes the sophistication of some of the voice recognition and lip-syncing tools available to distort photos and create videos. Last year, researchers from the University of Washington made public a disturbing project, in which they collected footage of Barack Obama’s lip movements and gestures and used it to make a realistic video of the former president lip-syncing a speech.
“If people can’t trust a service, they will stop using it,” says Turner. “And I think you’re already seeing that swing back to trusted news sources like The Australian Financial Review or the Wall Street Journal.”
At the same time, Turner says companies like Facebook exist because there is a need in the marketplace.
“I think they’ll respond and continue to exist but I also think you will see an emergence of decentralised models. And I personally believe there will be regulation as well as new approaches.”