As Facebook turns off facial recognition in Europe, is this the start of a change for the company?

Meet the new social network, not quite the same as the old social network.

After a long struggle with the Irish Data Protection Commission, Facebook is set to delete the last tranche of data kept from its facial recognition feature, dubbed Tag Suggestions, and turn it off for all users today. It just the latest retreat in a series of changes which may redefine the company.

The Tag Suggestions feature was first announced in December 2010. By using a mixture of information about facial shape and features, and contextual clues such as other people in the same album or picture, Facebook is able to suggest to users the names of other people in photos they have taken. Similar capabilities appear in other software – Apple's iPhoto, for instance, has an offline version – but Facebook's implementation leverages its vast user base to get more data than any competing company could manage.

However, Facebook implemented the Tag Suggest feature as an automatic opt-in for all users. That, combined with the fact that most photos on Facebook aren't uploaded by their subjects – obviously, since someone is normally behind the camera – meant that it necessarily played fast and loose with privacy concerns.

Just six months after it was announced, the first objections were raised in the US, and in August 2011, a Hamburg court became the first to rule that it must be opt-in to comply with local privacy laws. A month later, the Irish DPC began a wide-ranging privacy audit in response to complaints from a user group, Europe v Facebook, which included in its remit the facial recognition issues.

Since Facebook's European operations are based in Ireland – largely for tax reasons, since the company has a corporation tax rate of just 12.5 per cent for trading income – the decision of the DPC has wide-ranging effects. The first report, in December 2011, gave Facebook six months to comply with a number of requirements. "Shadow profiles" – profiles made of people who haven't joined Facebook from information uploaded by their friends – had to go, while data retention for searches and ad-clicks was limited, to six months and two years respectively.

The DPC also required Facebook to provide a prominent warning to its European users that it uses facial recognition technology that automatically tags them in photographs.

It was this last requirement which Facebook seems to have found too hard to comply with. In September, it closed Tag Suggestions to new users, and this month, it is shutting the feature entirely in Europe, and trashing the already collected data.

It's a bold move to take for a company which has, in other markets, been doubling down on facial recognition technology. In June, Facebook bought Israeli company Face.com, for a reported $55m. Face.com was the provider of much of the technology used by Facebook, and the company argued that the transaction "simply [brought]… a long-time technology vendor in house."

The company has always known that privacy concerns are one of the largest hurdles it has to to overcome. In its IPO prospectus, filed in February, Facebook highlighted a number of privacy-related risks to its business, from the publicity pitfalls associated with moving faster towards "frictionless sharing" than it's users are comfortable with, to the hurdles that stricter privacy regulation could introduce.

The facial recognition skirmish is an unusually under-the-radar battle for Facebook, however. Most of its highly publicised missteps involve public information being shared without the explicit permission or notification of users. This includes, for example, the ability of friends to "check in" people in Facebook Places without asking, as well as the various concerns over the frictionless sharing of social readers and apps like Spotify.

In fact, the first major privacy battle Facebook had to fight was over this type of issue, though in hindsight it demonstrates nothing so much as how much more comfortable we've become about sharing online. In September 2006, Facebook activated the News Feed, a feature now associated with the company more than anything other than, perhaps, the "like" button. But at the time, the idea of aggregating all this information – publicly available, but never before displayed in one place – was enough to spark user rebellion.

In what has become typical for Facebook, the company bet the business on people getting used to the new rules of the game. And they did, just like they did with the changed default privacy settings, the creation and promulgation of "@facebook.com" email addresses, and the aforementioned Places feature.

But three recent moves by Facebook suggest that the company may be changing its attitude, both voluntarily and as a forced reaction to circumstances.

The first is the deletion of facial recognition data, as well as the other changes mandated by the DPC. Facebook has always dealt quite well with user discontent – if only by successfully ignoring it – but when the law gets involved, it can be forced to backtrack far further than it normally would. It also means that it can be held to account for infractions of privacy which the average user simply won't notice.

Not many of us realised Facebook was even tracking search data, putting together a profile of us which we can't see, and few would have cared even if we did. But the DPC, like other information commissioners worldwide, has the authority and remit to ensure that data is collected with permission, and not retained indefinitely. Facebook knows it will face these problems with greater regularity as other nations step up to their responsibility to protect their users, and that will surely change its attitude.

The second is that Facebook itself has been backtracking from frictionless sharing, which had the potential to be one of the biggest clashes between it and its users. Andy Mitchell, Facebook's Manager of Media Partnerships, said last month that the company was moving away from it because user feedback wasn't good. This isn't just an issue with people being displeased that what they thought was private was in fact public – although that has happened as well.

For Facebook, the bigger issue is that the results of frictionless sharing just aren't particularly interesting. Sure, Facebook would like to know every news story you read, or every song you play, because it helps them build up a formidable picture of you to sell to advertisers. The problem is that social media is only interesting to anyone else if it allows people to present a curated vision of themselves. Nobody cares about the full list of songs you've played, but they may want to hear the one which is your absolute favourite at the moment. If Mitchell is to be believed, Facebook has come around to this way of thinking. The privacy benefits for users should be obvious.

The third change by Facebook is perhaps the most important. It is that the company is demonstrating a growing awareness that advertisement income alone cannot help the company achieve the goals its shareholders have set for it. It's tricky to estimate a price/earnings ratio for Facebook, since it hasn't released any results since it went public, but Business Insider estimate it's around 32. That means that you would need to hold Facebook stock for thirty-two years for it to make profit equivalent to the amount of capital you've provided them – or, more accurately, it means that the majority of Facebook's shareholders expect it to start making more money.

The problem is that Facebook's previous earnings growth has come largely from user growth. But with over a billion users, it starts to get very tricky to get any growth – the size of the planet is a constraining factor. As a result, Facebook needs to get more money per user.

One way to do this is, of course, to make ad space more valuable to advertisers, and that's what all of the company's social profiling is aimed at; but that's unlikely to be enough. For perhaps the best hint of the future, look to Facebook's recent launch of Facebook Gifts. The tagline is "Real moments. Real gifts." But perhaps the phrase "Real money" should be added there, because that's what is really important. Facebook wants you to spend real money buying gifts for friends through them – and then, of course, take a cut of the transaction that follows.

A Facebook which makes money from the services it provides, rather that providing services as a sidebar to its real business of selling your data to advertisers, is a company which has a vastly longer half-life. I hope they know that too.

The facebook hompage in 2005. Photograph: Wikimedia Commons

Alex Hern is a technology reporter for the Guardian. He was formerly staff writer at the New Statesman. You should follow Alex on Twitter.

Cleveland police
Show Hide image

Should Facebook face the heat for the Cleveland shooting video?

On Easter Sunday, a man now dubbed the “Facebook killer” shot and killed a grandfather before uploading footage of the murder to the social network. 

A murder suspect has committed suicide after he shot dead a grandfather seemingly at random last Sunday. Steve Stephens (pictured above), 37, was being hunted by police after he was suspected of killing Robert Godwin, 74, in Cleveland, Ohio.

The story has made international headlines not because of the murder in itself – in America, there are 12,000 gun homicides a year – but because a video of the shooting was uploaded to Facebook by the suspected killer, along with, moments later, a live-streamed confession.

After it emerged that Facebook took two hours to remove the footage of the shooting, the social network has come under fire and has promised to “do better” to make the site a “safe environment”. The site has launched a review of how it deals with violent content.

It’s hard to poke holes in Facebook’s official response – written by Justin Osofsky, its vice president of global operations – which at once acknowledges how difficult it would have been to do more, whilst simultaneously promising to do more anyway. In a timeline of events, Osofsky notes that the shooting video was not reported to Facebook until one hour and 45 minutes after it had been uploaded. A further 23 minutes after this, the suspect’s profile was disabled and the videos were no longer visible.

Despite this, the site has been condemned by many, with Reuters calling its response “bungled” and the two-hour response time prompting multiple headlines. Yet solutions are not as readily offered. Currently, the social network largely relies on its users to report offensive content, which is reviewed and removed by a team of humans – at present, artificial intelligence only generates around a third of reports that reach this team. The network is constantly working on implementing new algorithms and artificially intelligent solutions that can uphold its community standards, but at present there is simply no existing AI that can comb through Facebook’s one billion active users to immediately identify and remove a video of a murder.

The only solution, then, would be for Facebook to watch every second of every video – 100 million hours of which are watched every day on the site – before it goes live, a task daunting not only for its team, but for anyone concerned about global censorship. Of course Facebook should act as quickly as possible to remove harmful content (and of course Facebook shouldn’t call murder videos “content” in the first place) but does the site really deserve this much blame for the Cleveland killer?

To remove the blame from Facebook is not to deny that it is incredibly psychologically damaging to watch an auto-playing video of a murder. Nor should we lose sight of the fact that the act, as well as the name “Facebook killer” itself, could arguably inspire copycats. But we have to acknowledge the limits on what technology can do. Even if Facebook removed the video in three seconds, it is apparent that for thousands of users, the first impulse is to download and re-upload upsetting content rather than report it. This is evident in the fact that the victim’s grandson, Ryan, took to a different social network – Twitter – to ask people to stop sharing the video. It took nearly two hours for anyone to report the video to Facebook - it took seconds for people to download a copy for themselves and share it on.  

When we ignore these realities and beg Facebook to act, we embolden the moral crusade of surveillance. The UK government has a pattern of using tragedy to justify invasions into our privacy and security, most recently when home secretary Amber Rudd suggested that Whatsapp should remove its encryption after it emerged the Westminster attacker used the service. We cannot at once bemoan Facebook’s power in the world and simultaneously beg it to take total control. When you ask Facebook to review all of the content of all of its billions of users, you are asking for a God.

This is particularly undesirable in light of the good that shocking Facebook videos can do – however gruesome. Invaluable evidence is often provided in these clips, be they filmed by criminals themselves or their victims. When Philando Castile’s girlfriend Facebook live-streamed the aftermath of his shooting by a police officer during a traffic stop, it shed international light on police brutality in America and aided the charging of the officer in question. This clip would never have been seen if Facebook had total control of the videos uploaded to its site.  

We need to stop blaming Facebook for things it can’t yet change, when we should focus on things it can. In 2016, the site was criticised for: allowing racial discrimination via its targeted advertising; invading privacy with its facial-scanning; banning breast cancer-awareness videos; avoiding billions of dollars in tax; and tracking non-users activity across the web. Facebook should be under scrutiny for its repeated violations of its users’ privacy, not for hosting violent content – a criticism that will just give the site an excuse to violate people's privacy even further.

No one blames cars for the recent spate of vehicular terrorist attacks in Europe, and no one should blame Facebook for the Cleveland killer. Ultimately, we should accept that the social network is just a vehicle. The one to blame is the person driving.

If you have accidentally viewed upsetting and/or violent footage on social media that has affected you, call the Samaritans helpline on  116 123 or email jo@samaritans.org

Amelia Tait is a technology and digital culture writer at the New Statesman.

0800 7318496