On the morning of 28 October last year, the day of Iceland’s parliamentary elections, Heiðdís Lilja Magnúsdóttir, a lawyer living in a small town in the north of the country, opened Facebook on her laptop. At the top of her newsfeed, where friends’ recent posts would usually appear, was a box highlighted in light blue. On the left of the box was a button, similar in style to the familiar thumb of the “like” button, but here it was a hand putting a ballot in a slot. “Today is Election Day!” was the accompanying exclamation, in English. And underneath: “Find out where to vote, and share that you voted.” Under that was smaller print saying that 61 people had already voted. Heiðdís took a screenshot and posted it on her own Facebook profile feed, asking: “I’m a little curious! Did everyone get this message in their newsfeed this morning?”
In Reykjavik, 120 miles south, Elfa Ýr Gylfadóttir glanced at her phone and saw Heiðdís’s post. Elfa is director of the Icelandic Media Commission, and Heiðdís’s boss. The Media Commission regulates, for example, age ratings for movies and video games, and is a part of Iceland’s Ministry of Education. Elfa wondered why she hadn’t received the same voting message. She asked her husband to check his feed, and there was the button. Elfa was alarmed. Why wasn’t it being shown to everyone? Might it have something to do with different users’ political attitudes? Was everything right and proper with this election?
Iceland had just reached the end of its most arduous and dirty campaign season ever. For weeks, anonymous accounts had been spreading accusations on social media about nearly every political candidate. The country had been flooded with bizarre “exposé” videos on Facebook and YouTube. Some had been viewed millions of times, even though Iceland has only around 340,000 residents. And now this button. Was there a connection?
Elfa watched as more and more people responded to Heiðdís’s post. Some had seen the button, others had not. Out and about on election day, Elfa asked everyone she met about it. It was clear that not everyone had received the message. Of those that did, some got it later than others. Some had seen it while scrolling through their Facebook newsfeeds; for others it appeared at the top. Meanwhile, responses to Heiðdís’s post appeared to show that users given the button option were not randomly selected either. Minors and non-citizens were not shown it: only those in the voting population. But then, as she was also discovering, not all of them. Was there some kind of pattern?
Immediately after the elections were over, the button disappeared. There was no sign of it on Facebook’s company site Newsroom or on any government sites. Elfa rang a friend, Kristín Edwald, chair of the Icelandic Election Commission. Edwald was totally surprised; she had never heard about this button. Not even the special commission, given the job of working on an update to the election law, knew what Elfa was talking about.
Facebook’s ‘get out the vote’ button made an appearance in the 2014 Scottish independence referendum.
Elfa had an idea why none of the authorities had taken the button seriously: at first glance, Facebook’s get-out-the-vote button seems harmless. What effect could a simple election day reminder possibly have?
The Icelandic elections are simply the most recent time the button has been employed for major elections in the west. In the US, it was first used – and fully disclosed by Facebook – in 2008, and again in 2010 and 2012. Facebook has published its own studies about its effects. Initially, despite some scepticism on the left, it was mostly seen as a positive tool, bringing people to the ballot. Facebook was an emerging phenomenon, with a couple of hundred million users.
The first known cases of the button being used outside the US are the Scottish referendum in 2014, the Irish referendum in 2015 and the UK election later that year, all of which were communicated by Facebook. After that there was silence about the button – and no further public statements by Facebook. But the company has now revealed, in answer to questions in the preparation of this article, that the button was used in the UK in the 2016 European Union referendum, the 2016 US presidential election that brought Trump to power, and in Germany’s 2017 federal elections.
But what effect did it have? That we don’t know. And if Facebook does, it’s not saying. Did its button make a difference in crucial, closely fought recent elections? In the EU referendum in Britain or the election of Trump in the US?
In the ongoing debate about the influence of Facebook on politics, the issues largely revolve around third-party applications. The crucial distinction about the vote button is that it is made and operated by Facebook alone: this is Facebook itself becoming a political actor. Indeed, in his testimony to the US Congress, Facebook chief Mark Zuckerberg proudly cited the use of the button during the last US presidential election.
And this was what concerned Elfa in Iceland. The main reason for her worries was a presentation she had seen two weeks before the Icelandic elections, at a conference of the Council of Europe and the Organisation for Economic Co-operation and Development in Vienna. It had been part of an OECD panel on fake news, Russian manipulation online and personalised advertising. At the heart of the matter was abuse of social networks by third parties. The button, however, was something else entirely. It was Facebook’s very own tool. And, in contrast to all the speculation around the actual effects of fake news, Facebook had numbers.
The presentation had been given by the Austrian digital ambassador to the EU, Ingrid Brodnig. Brodnig had spoken about one of the largest ever experiments in the field of social science. It had taken place on the day of the 2010 US congressional elections, when Facebook suddenly sent a voting reminder to 61 million US users – a quarter of the US voting population. Every US Facebook user over the age of 18 who logged in on election day saw the message. Then a team from Facebook, together with researchers from the University of California at San Diego, weighed Facebook datasets against election returns. The goal was to find out whether the “voter button” actually got out the vote.
The results of the study were published in September 2012 in the journal Nature. The conclusion: the button works. Users who saw it were more likely to vote. The effect was slight, but scaleable in the millions at just the cost of a line of code. It made the button the most effective voter activation tool ever built, creating 340,000 additional voters. It was possible, concluded the researchers, that “more of the 0.6% growth in turnout between 2006 and 2010 might have been caused by a single message on Facebook”.
If Facebook had shown the button to every US voter, more than a million voters could have been mobilised. This was the first scientific evidence of the real influence that the still-new network could have – an important message for Facebook’s potential advertising customers. Elfa heard all of this for the first time at that conference in Vienna. The Austrian digital ambassador concluded her speech by saying: “There has never before been so much power in the hands of a single company.”
In 2012, Facebook once again tested its capacity for influence. This time in the US presidential elections. It was announced on Newsroom that all users would be shown the button (which turned out not to be the case). The results were published (by Facebook – there was no independent study) in April 2017 in the science journal Plos One. The button had worked again: 270,000 additional votes were cast.
For users who got both the button as well as notifications from friends who had voted, the rise in participation was 0.24% this time. As slim as that seems, in 2000 George W Bush beat his presidential opponent Al Gore in the decisive state of Florida by 537 votes. That’s 0.01%.
On US election day in 2012, media reports indicate that Facebook did other tests to optimise the button. Even today not much is known about them. Facebook has never revealed publicly how many variations of the button they tested. But the company apparently wanted to know which worked better: when Facebook simply displayed the button or when it came as a recommendation from a friend.
This was the reason only some people saw the election day reminder at the top of their newsfeeds while others saw it as a post shared by a friend. Some people saw the button only at their computers, others saw it on all devices (a Facebook project manager at the time, Antonio Martínez, says that in 2012 he heard from co-workers that developers were undecided whether to show the button on iPhones: that alone could bias election results, they worried, because iPhone users tended to be more urban and liberal). There were multiple variations in the text; some read “I’m a voter”, while others came up “I voted”.
The key questions are: how much of a difference does it make when Facebook turns on its button? And could Facebook potentially distort election results simply by increasing voter participation among only a certain group of voters – namely, Facebook users? At its core, Facebook is an ongoing experiment being conducted on society. In Facebook’s eyes, we are all the subjects of a global experiment in profit-maximisation. No one can predict precisely what the effects of a certain program alteration will be, so everything is constantly being tested.
Facebook can see in great detail how we react to every single alteration they make. The Facebook algorithm that’s always being talked about does not exist. It’s more accurate to say there are many continuously-under-development programmatic threads that interact to determine what comes up in our individual Facebook feeds. The aim is always to increase “engagement” – the time we spend interacting with the platform. For our part, we notice that we are being experimented on just about as much as rats in a maze do.
Only once has Facebook apologised for an experiment: in 2014 it was revealed that it had tested 689,003 users to determine how much their feelings could be influenced. During the “emotional contagion“ experiment for one test group, positive posts from friends were partially withheld; for the other group, negative ones. Even if the effects were slight, Facebook showed itself to be a sort of thermostat, with which the moods of its users could be regulated. The very thought of it: a company which, in order to test its product, risks the psychological wellbeing of its users – this created an uproar. But one thing Facebook learned from the backlash was that it was better keeping things secret.
When something problematic does get revealed, Facebook turns to the same arguments: it’s the fault of ill-intentioned, careless or clueless users. In the case of Cambridge Analytica, Facebook first blamed the users whose data had been misused for being unable to comprehend its terms of use policy. Then they cast Cambridge Analytica as the villains who had taken advantage of user ignorance. Facebook is always the victim, never the culprit.
The Facebook employee who first exported the “voter button” outside the US is the London-based Californian Elizabeth Linder. She is 34, a Princeton graduate and was an early recruit at Facebook, starting in 2008 (she left in April 2016). In an upmarket vegan restaurant in central London, she relates how she built up Facebook’s “global politics and government outreach” department. Working from her London office, her task was to persuade the political classes of Europe, the Middle East and Africa that Facebook was the place where voters were – and that, therefore, politics has to be there too. At first it was rough going. “The politicians all thought Facebook was just something for young people,” she said.
Then came the Arab spring. Young activists used Facebook to network, spread their ideas and organise demonstrations. Governments in north Africa were toppled. The Arab spring was the best marketing for Facebook ever. It turned a toy into a tool of power. Politicians and government officials from all over were busy trying to contact Facebook, and it was Linder they spoke to. She has advised the Vatican and British parliament; she brought the Dutch royal family on to Facebook. She says she spent the eve of the second Tahrir uprising with the social media department of the then Egyptian president Mohamed Morsi. She had one clear aim: “Fulfilling our mission to make the world a better place.” In Macedonia, she advised pro-Europe organisations; in Lithuania, Radio Free Europe; in Uganda, women’s groups.
“The first time I used the ‘voter button’ was in Scotland,” says Linder. “That was 2014, the independence referendum.” She doesn’t see anything wrong with the button. It is an excellent reputation marker, something that proves the impact of Facebook, she says. “The argument I used to convince Facebook [to extend use of the button globally],” says Linder, “was the whole thing about the button really having an impact on voter participation in America, and how we could achieve the same impact internationally.” After all, by then 70% of Facebook users lived outside the US. The button wasn’t some business scheme, says Linder. You can’t make money with it.
Did Facebook have a bigger plan? She says not. Were there concerns about potential political influence? Only occasionally. One example she mentions is Lebanon, where she avoided giving political consultation, not wanting to accidentally advise people who could be on a terrorist watchlist. Who determined when the button should be deployed? Her alone, Linder says. The Scottish independence referendum was, in her view, a good place to start: only two camps, similar to the US. “The second one, I think, was a referendum in Ireland.” Then the 2015 elections in Britain. The long-term goal was to use the button in “every major election”. In Britain in 2015, Facebook made an effort to communicate it, did a lot of marketing. Later, they wouldn’t.
And yet there is indeed one public announcement of Facebook’s plans to implement the “voter button” worldwide, from 2014. At that time, the project was publicly well received. Since then, however, Facebook has been silent on the matter. There is no data or any accounting available about where the button has been deployed, except for one announcement in India.
Searches for the button on the internet yield private screenshots posted from many places, including India, Colombia, Holland, Ireland, Hong Kong, South Africa – even from the European parliamentary elections. Also represented are countries in difficult geopolitical environments, such as South Korea and Israel, and endangered democracies such as the Philippines, Turkey and Hungary.
What remains unclear about the button – this powerful tool that a US corporation can insert into elections globally completely free of public scrutiny – is when and why it gets deployed. Do you have to pay to have it activated? Or is it enough simply to have a cordial relationship with a Facebook employee?
“As far as I know, there is no public list of where the ‘voter button’ has been used,” says Linder. And within the company? “Yes.”
She left Facebook in 2016. Why? “I wanted to stay in Europe,” she says simply. When pressed, she shrugs. It had always been her dream to live in London. And, no, she has never been to Iceland.
When I approached Facebook to ask why they had used the button in Iceland, a public relations firm replied: “We show a message on election day to remind people to vote.” Its explanation for the button not being shown to every user had to do with users’ individual notification settings or their use of an older version of the app. As to how strongly voter participation had been influenced in Iceland, there was no comment. Nor did Facebook say who was actually shown the button and who was not. That information is “unfortunately” not available “for any country”. The list of countries in which the button had been used could not be provided either.
Why doesn’t the company want to reveal this information? To what extent has the button influenced election results in recent years? What political data is being gathered? Are models being tested and strengthened? Voter participation in Iceland turned out to be surprisingly high. There were increases among both young and old, even though experts had been speaking before the poll of a general sense of election exhaustion. Whether this was due to the button, and who benefited most from the additional votes, is impossible to tell without more information from Facebook.
Facebook is a US company without an office in Iceland. In a certain way, its use of the button constitutes interference by a foreign actor. Facebook has meddled in the democratic elections of a foreign country, and no one outside of the company seems to know anything about it. Replies to inquiries with Iceland’s justice ministry, election commission and intelligence services indicate that they were ignorant on the subject. Even Páll Þórhallsson, the chief legal counsel to the prime minister and an internationally experienced media lawyer, had not been informed. But the deployment of the button was probably not illegal under Icelandic law. Legislators could hardly have predicted these kinds of methods, so why formulate any legal text to cover them?
Facebook had sent a team to Iceland, on 10 October 2017, just 18 days before the election. In a parliamentary meeting room – accessible only with a pass – two representatives, Anika Geisel of Facebook’s politics and government outreach team and Janne Elvelid, the company’s Stockholm rep, had been meeting leaders from the major parties. They had given a two-hour introduction on the ways politicians could engage on Facebook with potential voters – what they could achieve with the platform, how they could maximise the “engagement” of their fans. The examples they showed were in German. Facebook cited the page of Sahra Wagenknecht, of Germany’s far-left party die Linke, as well as the fan page of then soon-to-be Austrian chancellor Sebastian Kurz, as examples of “best practices”. It was a standard promotional presentation, with examples pulled from the Austrian and German elections – the slides were in German.
But why was Facebook in the country at all? The tips their team gave were banal. The presentation left most visitors wondering why the meeting had been held – there was no mention of paying Facebook or of advertising opportunities. Nor were there any answers given to the pressing questions on who was behind the wave of fake news videos that had flooded the country. The “voter button” wasn’t mentioned. Facebook had been invited by the conservative Independence party, which declined to comment for this article. Elfa, who speaks Swedish, decided to email Janne Elvelid. At first he responded warmly but when she sent him questions about the button, he suddenly had no more time for written communications. She should call him, he said.
So on 8 February this year she did call him. He had been assigned to Iceland, he told her. He had discussed deploying the “button” beforehand, by telephone, with someone in Iceland’s justice ministry. Facebook had sent the button to every user, he said. Still, Facebook can’t really tell who actually sees the button; algorithms and user settings would ultimately determine that. Facebook cannot control who deactivates which notifications.
In statement after statement, Facebook insists it supports democracy. But in Iceland, after six elections in one decade, with confidence in the political system shaking, and when democracy needs nothing more desperately than trust, Elfa was left with nothing but doubt.
The experiment is still going on. Early last month, an acquaintance sent Elfa a screenshot from the elections in Italy. The button showed up there too. In response to my questions, Facebook stated that it had also deployed the button in the last federal and regional elections in Germany as well as for the Brexit referendum.
In Iceland, outrage about Facebook is beginning to mount. Meetings are being held, including with MPs and the prime minister’s office. Elfa is still wondering why Facebook is investing energy in introducing its button all over the world. The only explanation that makes sense to her is that elections help to bring people to the platform; that users, as soon as they see that their friends voted, will take their political battles to Facebook, and, with their increased engagement, help Facebook bring in more advertising. That would be the profit. The cost is borne by democracy.