It’s good to talk (Image: Bruno Mangyoku)
AS ARTURO BEJAR clicked through the millions of photos it was hard to see what was wrong with them. Happy couples, cute puppies, every now and then a person caught mid-chew. Yet each had upset someone enough to ask Facebook to delete it.
Bejar could hazard guesses for the reasons behind some of the requests – maybe this couple had broken up, maybe that guy was embarrassed by his double chin. But the person who reported the photo was usually in it and the person who posted it a friend. People were asking Facebook to intervene in personal grievances. Why not work it out themselves? One common complaint was people posting pictures of other people’s children. For Bejar, Facebook’s director of engineering, it was as if your photographer brother-in-law decided to exhibit photos of your kids. Would you ask the gallery to take down the photos instead of talking to your brother-in-law?
Advertisement
We tend to think that technology has magical properties that change who we are, says psychologist at the University of California, Berkeley. But that’s not the case. “Facebook has all the problems that real life has. It’s still just human society.”
What was needed was a way for people to interact online more like they do offline. Bejar put a team together to bring compassion to our online interactions. With more users than the population of China, if Facebook could get people to talk things out, it would establish a new standard for how we behave on the internet. Compassion could go viral.
Most social media platforms, from Facebook and Twitter to YouTube and Vine, have some kind of reporting or “flagging” system to manage content. The monitoring of flagged material is carried out by moderators trained to enforce the site’s community standards. In Facebook’s case, that means depictions of violence or nudity get deleted, as do posts containing hate speech.
Policing, though, has obvious limitations. What counts as acceptable can depend on context, which varies between cultures and social groups. , Facebook is unlikely to get it right all of the time. And since policing puts moderators in the role of a teacher meting out justice, it is also a lot of work.
Bejar wanted to do better. By revamping Facebook’s system for flagging posts, he hoped not only to cut down the millions of photos to be sifted through each week but also to improve people’s experience of using the site. But that’s not as simple as it sounds. One of the biggest challenges to online communication is typing. “People don’t realise the written word is quite different from the spoken word,” says Keltner, who is on Bejar’s team. With text, we lose out on cues such as eye contact, touch and tone of voice that convey a lot of information.
In 2005, for example, at New York University and colleagues found that – yet we still believe we are being perfectly clear. People greatly overestimate their ability to communicate clearly online.
What’s more, the internet brings out the Mr Hyde in all of us. Psychologists have long noted how . at Rider University in Lawrenceville, New Jersey, has suggested that . “There’s been an exponential rise in a form of communication that we’re not yet equipped to handle,” says Keltner.
“There’s been an exponential rise in a form of communication we’re not yet equipped to handle”
Friendly feedback
In 2012, Bejar assembled a team of engineers and psychologists at Facebook’s headquarters in Menlo Park, California. At the time, Facebook offered people three choices for dealing with a friend’s post they found upsetting: report content that violated a Facebook policy, hide it from their personal view or “unfriend” the person.
It was a limited palette. “There was no opportunity for the feedback – the disdain or disapproval, the recognition or approval – that a person would have felt in a face-to-face conversation,” says social neuroscientist at the University of California, Berkeley. The commonly chosen option to unfriend the other person killed the conversation. And reporting relied on Facebook to act, which it could only do if the content violated one of its policies.
“We needed to make it easy for users to explain what they were feeling instead of relying on an authority to take care of things,” says Simon-Thomas. “This may all seem obvious, but it shows you how much of a chasm there has been between the fields of technology and social science.”
The team added an option so people could contact others who had posted a photo they wanted removed. At first the team provided a blank box in which to compose a message, but only 20 per cent of people used it. So they filled the box with a default response. Simply adding the words “Hey, I don’t like this photo. Please remove it” got more people using the system (see “A way with words”).
Then they started tweaking. Letting people select why they didn’t like a photo from a list that included options like “It makes me sad”, “It’s embarrassing” or “It’s a bad photo of me” bumped up engagement even more. By giving people prewritten text that added emotional context, the rate of one-to-one interaction tripled. Once they started experimenting, the team found that minor tweaks made big differences. With more than 1.4 billion active monthly users, Facebook was affecting a whole lot of people.
Softening our written words was key. We tend to read the worst into online messages because they offer so few emotional cues. Research by at Syracuse University in New York suggests this negative effect is amplified when we read messages from someone of higher status, such as our boss. He found a difference between the sexes too: , for example.
“This is one of the reasons why sarcasm is so hard to get across,” says Simon-Thomas. “No one ever gets the smirk.” To help people get the joke, the team designed emojis – small images of faces – to allow an even broader range of expression. We may be learning to process such images emotionally. A recent Australian study found that .
Smiling may be universal, but many things aren’t. When Facebook’s dispute process began reflecting cultural differences, it doubled the number of people choosing to resolve issues themselves. In India, where making fun of a favourite celebrity is deeply offensive, people can choose “This photo insults someone important to me”. In Turkey, people can opt for “This promotes militant activities, violence, armed attacks or weapons”.
For teenagers, the system needed to be adapted further. After leading teen focus groups, , a Yale psychologist on the team who specialises in emotional intelligence, advised changing “Report this post”, which to many implied getting a friend into legal trouble, to “This post is a problem”.
“Kids took from the language that this was a generic experience that didn’t apply to them,” Brackett says. “We avoided words like ‘stalking’ and ‘harassed’, and instead used ‘threatened’ or ‘bothered’, to make the experience much more conversational and applicable to them.”
Depending on how embarrassed, sad or scared someone is, Facebook now offers different solutions. A user who is annoyed by someone’s post – but doesn’t necessarily want it removed – is given a separate option simply to tell that person how they feel. Teens may also be given a link to a bullying hotline and other counselling resources. There are guides for parents, teachers and even the bullies to help them understand what they can do better online and in real life. All of these changes doubled the number of teens seeking help or contacting the person who posted something hurtful, Brackett says.
Facebook is extending this direct and emotionally engaging approach across its business. In its privacy controls, for example, you now see “Who can see my stuff?” and “How can I stop someone from bothering me?” instead of legalese.
Facebook is also interested in looking at how anonymity can be used to communicate in a positive way. Rob Boyle, a product manager, says he can see potential in using anonymous signals to kick-start difficult conversations – like when you want to gently break it to a friend that posting about politics every 10 minutes is getting annoying. “We’re looking to build more ways to convey subtle signals, without making it too easy to slide into the negative,” says Boyle. “It’s a tricky balance, because the effects go both ways. That’s why we’ve never done the dislike button.”
This was a lesson learned by Disqus, which provides the discussion platform for the 1 billion monthly users of WordPress, Tumblr and many other web sites. Disqus has been looking at how voting comments up or down affects conversations. CEO Daniel Ha is excited about Facebook’s work because it gets people to stop and think. “As communication platforms, it should be our intention to get the best out of people rather than just their first reactions,” he says.
Thumbs up, thumbs down
Disqus added voting as part of a scoring system that moves the best comments to the top and alerts moderators to people who repeatedly post comments that get downvoted. In 2012, Disqus introduced a requirement that people log in to vote, and began displaying their name and profile picture next to comments they voted up.
Making votes more visible amplified the positive signals in the communities, says Ha. Reports of abuse fell by almost 80 per cent. But the change also began to encourage negative feedback. “We saw retaliation and vindictive trolling, people ganging up on each other,” says Ha. “People would obsess over who had voted them down.”
Justin Cheng at Stanford University and colleagues looked at Disqus’s data and found that positive feedback had no effect on what people posted, but downvotes made users more likely to post antisocial content. .
The solution, Disqus decided last year, was to hide all downvoting statistics, using them only as a background measure of quality. “We’re seeing less noise and frustration,” says Ha. “People are replying to each other, rather than just registering another downvote.”
Del Harvey, Twitter’s head of Trust & Safety, has also been following Facebook’s work with interest. On Twitter, where fewer people know each other, building channels for direct communication isn’t the best option. Instead, the solution may involve giving people more control over what they see, says Harvey. For example, Twitter may explore a form of community reporting in which people in your chosen network could rate tweets, and enough downvotes would hide the content from your view. New accounts that use language similar to previously flagged tweets are also filtered out. In addition, Twitter has tripled the number of staff policing its content.
By its own estimation, Facebook’s efforts to get us to behave better online have been a success – more than 8 million people use the social resolution tools each week to deal with conflicts over photos, status updates and shared links. For Facebook, the ideal outcome isn’t necessarily the removal of the photo or post, which might be valuable to the poster. It is opening a dialogue.
The team tried to adapt the platform so that an emotionally intelligent response is the easiest thing to do. When asked to remove a photo, people do so more than half of the time, and around 75 per cent at least write back. Even if a handful of these replies are negative, the simple acknowledgement often helps people to get over their initial reaction.
Facebook may have helped millions of people express themselves, but some will feel uncomfortable with yet another example of subtle manipulation (see “Crowd control“). However, the internet calls for special measures. Bejar says he started thinking about compassion technology because of his son. “I had set a personal deadline,” he says. “I’d need to feel really good about the content on Facebook by the time he got an account.”
“Now he is almost at that age, we’re almost there,” he says. “The broader internet is a different story.”
Leader: “It’s time we turned social networks into real communities“
Crowd control
Subtle tweaks to Facebook’s algorithms can have seismic effects. Last year, for example, there was outrage over the ethics of treating Facebook users as guinea pigs in a study looking at the effects of its newsfeed algorithms. But Facebook has always tweaked what we see. “This was part of ongoing research companies do to test different products,” COO Sheryl Sandberg said at the time. “We never meant to upset you.”
Facebook is also out to change how we behave (see main story). For example, during the 2010 presidential election in the US and the recent general election in the UK, an “I voted” button let users tell others when they had cast their ballot. Robert Bond at the University of California, San Diego and colleagues estimated that US voter turnout in 2010 by around 340,000.
And in February, Facebook announced its “Speak Panzagar” campaign to promote civility in Burma. It launched cartoons with positive slogans that can be shared between the country’s 600,000 Facebook users. Panzagar, which means “peaceful speech”, is a grassroots effort in Burma to counter hate speech aimed at the country’s Muslim minority. Slogans include “Think before you share” and “Don’t believe it”. According to Facebook, the most widely shared “sticker” – illustrations or animations sent between Facebook friends – in Burma, and proved “an extremely powerful way to promote civility”.
This article appeared in print under the headline “It’s good to talk”
