For years people have clamored for Facebook to add a “dislike” button alongside its iconic like button. That was never going to happen, for reasons I’ve explained. Instead, in October the company began testing a suite of six emoji complements to the like button: “love,” “haha,” “yay,” “wow,” “sad,” and “angry.”
On Wednesday, it rolled out the new buttons to Facebook users worldwide, minus one: “yay.” (“Yay” missed the cut, Facebook product manager Sammi Krug told me, because the company’s testing revealed that people found it vague and didn’t use it as much.)
Facebook explained the new feature, called Reactions, in a blog post:
We’ve been listening to people and know that there should be more ways to easily and quickly express how something you see in News Feed makes you feel. That’s why today we are launching Reactions, an extension of the Like button, to give you more ways to share your reaction to a post in a quick and easy way.
You can see the new buttons by either holding down the like button (if you’re on your phone or tablet) or hovering over it (if you’re on your desktop). Tap the emoji reaction you want, and the icon for it will appear beneath the post, just as the like icon does.
Facebook will tell you that the new reactions are all about giving users new ways to communicate and express themselves. No doubt that’s part of it. Users have long complained that like doesn’t feel appropriate in a lot of circumstances, such as when a friend’s loved one has died or an acquaintance posts a political screed that you find interesting but also troubling.
But, like almost everything Facebook does, there is a double purpose at work here—and that second purpose involves data. Specifically, Facebook is now going to be able to collect, and profit from, a whole lot more of it.
In a January Slate cover story, I looked behind the scenes at how Facebook’s news feed algorithm works—how it decides what you see at the top of your feed every time you log in—and why the company keeps tweaking it. In short, Facebook has come to believe that the key to its long-term success lies in gathering ever more and ever richer data on how its users react to the posts they see in their feed. The company can use that data to personalize each user’s feed to her liking, so that it never becomes so stale, repetitive, or overwhelming that she’s tempted to look elsewhere for her daily fix of updates from friends. Much of the same data goes into the software Facebook uses to decide which ads its users see in their feeds.
In contrast, giving users six reaction options means that Facebook can start to gather much more nuanced data on how users are reacting to any given post. It can begin to differentiate between posts that users are enjoying, posts they find fascinating, posts that make them happy, and posts that make them sad.
Facebook says that it isn’t using the data from new reactions in that way—yet. But that will soon change, according to a blog post from Facebook’s Krug:
Initially, just as we do when someone likes a post, if someone uses a Reaction, we will infer they want to see more of that type of post. In the beginning, it won’t matter if someone likes, “wows” or “sads” a post — we will initially use any Reaction similar to a Like to infer that you want to see more of that type of content. Over time we hope to learn how the different Reactions should be weighted differently by News Feed to do a better job of showing everyone the stories they most want to see.
How might that work? In our interview, Krug declined to get specific, saying the company’s goal for now is just to learn about how people are using the new reactions so it can improve the feature over time. But I have some educated guesses.
Here’s one idea: Facebook’s own research has shown that its news feed algorithm can make users happy or sad by showing them a greater proportion of positive or negative posts. Further research might help the company figure out the optimal mix of happy, sad, amazing, funny, and infuriating posts to keep users coming back to the news feed every day. Think of it as a modern-day spin on the way media editors have always aimed to offer a mix of hard news, human-interest features, and entertainment on their front pages or in their newscasts. Over time Facebook could even adjust that mix for each user. So people who come to Facebook looking to be amused would find their news feeds packed with funny posts. Those who come looking to be informed, challenged, or provoked would find their feeds peppered with news stories and controversial opinions.
Reactions data could be just as potent when applied to Facebook’s advertising algorithms. Advertisers have always been keen to understand how people are reacting to their ads. Reaction buttons, applied to Facebook’s massive audience, could amount to a valuable analytics tool for them—and for Facebook, which places a high priority on showing its users ads that they find worthwhile.
Advertisers have also always been concerned with the editorial context amid which their ads will appear. In the future, Facebook might be able to ensure that their ads come sandwiched between posts that are making people happy or that they never appear alongside posts that are making people angry.
But why only six reactions? Versatile as they may be, there’s no way that “like,” “love,” “haha,” “wow,” “sad,” and “angry” can adequately cover the spectrum of human emotion. Besides, a lot of other social networks and messaging services—including Line, WeChat, Slack, and even Apple’s iMessage—give their users access to hundreds of different reaction emoji. Why won’t Facebook?
One reason might be that Facebook has found in its testing that users are more likely to choose an alternative to the like button if the range of alternatives is small enough to quickly grasp. The paradox of choicesuggests that more options might simply overwhelm people. That’s probably a big part of why it dropped “yay.”
A deeper reason, however, may be that the data from users’ reactions will be better structured, and therefore easier to incorporate into Facebook’s algorithms, if there are only a handful of discrete options. After all, people can already react to posts in an infinite variety of ways, simply by commenting on them. And Facebook does have machine-learning engineers working hard on natural language understanding algorithms to help decipher the sentiments that underlie those comments. But that’s a seriously difficult task.
With the reaction buttons, Facebook’s engineers won’t have to do the hard work of classifying users’ reactions to a given post—because the users will be doing it for them.