How Images Become Viral on Google+
Network science has changed the way we think about the spread of information, diseases and even fashions. Perhaps its most important revelation is that the connectivity of the network matters when it comes to the viral spread of almost everything.
For example, the eventual size of a forest fire is determined by the connectivity of trees—how close they are together, for instance—but it is almost nothing to do with the size of the spark that started the fire in the first place.
The same idea explains why some internet messages become viral while others that seem just as interesting or funny or outrageous, never get anywhere. It’s all to do with the state and connectivity of the network at the moment the message is released.
And yet there is surely some important characteristic of the message content itself that makes people want to pass it on. In recent years, various research teams have studied the content of viral messages to see what kind of magic sauce they have in common.
For example, one study looked at articles published in the New York Times and the emotions evoked in the comments section. It found a clear relationship between the strength of feeling and the virality of the article.
Strangely, little or no work has been done on the virality of images, despite the hugely important role they play on the Internet. Today that changes thanks to the work of Marco Guerini at Trento RISE in Italy and a couple of amici.
These guys have studied the characteristics of images that spread virally on Google+. And they’ve discovered a number of common features about these images and how they appeal to people in different ways.
Guerini and co began by harvesting almost 300,000 public posts from the top 1000 most-followed users in Google+ between June 2011 to June 2012. Of these, some 175,000 contained static images, 13,000 contained animated images and 100,000 were text-only posts.
The team studied three indices of virality–they looked at the way each of these posts was reshared, plusoned (equivalent to being liked) and replied to.
And they divided the images into various categories to see how each fared. For example, they compared the virality of images versus text-only posts, finding that posts with an image are significantly more likely to be re-shared, something perhaps to be expected.
But they also found intriguing effects. Posts with fewer than 75 plusoners are more likely to contain images whereas posts with more than 75 plusoners are more likely to be text only. Guerini and co think they know why. “While it is easier to impress with images in the information ﬂow…high quality textual content can impress more,” they say.
Another comparison they study is between static and animated images. The latter are significantly more likely to be re-shared while static images are more likely to be replied to or plusoned. Guerini and co say the higher rate of re-sharing for animated images can be explained because animated images are often designed to convey “a small “memetic” clip”.
Curiously, vertical images tend to be more viral and horizontal ones. Guerini and co put this down to the fact that vertical images are more likely to contain a portrait of a celebrity than horizontal images. (Most of the most-followed user son Google+ are celebrities of one kind or another.)
These guys also look at black and white versus colour images, bright images versus not so bright images, and so on.
Finally, they looked at how their virality indexes are correlated with each other. They say that the plusone and replied to indices are highly correlated while the reshared index appears to be independent of these. “We hypothesize that plusoners and replies can be considered as a form of endorsement, while reshares are a form of self-representation,” they say.
In other words, most users consider the act of resharing as a form of self-expression that says something about themselves. Whereas replying and plusoning play a different role.
That’s an interesting first step in exploring the virality of images. And there is clearly work ahead. Guerini and co say they next want to look at factors such as the composition of a picture and its content using scene/object recognition.
Whether this kind of work will ever help celebrities (or anyone else) make their posts more viral is open to debate. But it clearly won’t stop them trying.
Ref: arxiv.org/abs/1309.3908: Exploring Image Virality in Google Plus
Geoffrey Hinton tells us why he’s now scared of the tech he helped build
“I have suddenly switched my views on whether these things are going to be more intelligent than us.”
ChatGPT is going to change education, not destroy it
The narrative around cheating students doesn’t tell the whole story. Meet the teachers who think generative AI could actually make learning better.
Meet the people who use Notion to plan their whole lives
The workplace tool’s appeal extends far beyond organizing work projects. Many users find it’s just as useful for managing their free time.
Learning to code isn’t enough
Historically, learn-to-code efforts have provided opportunities for the few, but new efforts are aiming to be inclusive.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.