The two-tone dress (left) alongside an ivory and black version, made by Roman Originals.

Photographer: Joe Giddens/PA Wire/AP Photo

It's Not About 'The Dress'

Megan McArdle is a Bloomberg View columnist. She wrote for the Daily Beast, Newsweek, the Atlantic and the Economist and founded the blog Asymmetrical Information. She is the author of "“The Up Side of Down: Why Failing Well Is the Key to Success.”
Read More.
a | A

Twenty-eight million times.

As of this afternoon, that’s how many times people had viewed Buzzfeed’s poll on the infamous dress -- you know, the one that’s white and gold, unless it’s blue and black. Depending on what part of the picture I look at, I’ve seen it both ways. And as I stared at the Internet’s favorite optical illusion, I’ve been pondering the mystery of viral content.

My most trafficked post ever was this quickie, which I wrote while traveling, squatting on the floor of an airport with my laptop balanced on my knees. I think it took me 15 minutes, most of which was spent searching Google Books for evidence that Martin Luther King had actually given voice to a quotation that was, at that moment, going viral on Twitter and Facebook. He hadn’t, but the fake quote got millions of posts and retweets -- and every time someone posted or tweeted it, someone else replied with my skeptical take. Hello, Internet fame.

Needless to say, if you had asked me while I was sitting in the airport if you thought this post had a shot at being the most-trafficked thing I would ever write, I would have laughed. I have an interest in fake quotations, and I had a few minutes between flights, so I thought of it as a fun squib to share with my readers, not as a potential megahit. Also needless to say, I was wrong.

If you believe Duncan Watts, a sociologist who works for Microsoft Research, that’s no accident. In his terrific book "Everything Is Obvious (Once You Know The Answer)," he argues that popularity really does have a significantly random component. For example, he constructed an experiment, which I describe in my own book, to determine how songs become popular in different social networks. If you randomly assigned people to groups that could listen to music and display their preferences, would groups converge upon the same songs as most popular, or would they each pick their own sets of “best” and “worst”?

The answer is that there was no clear “best” song; instead, each group picked their own best. It doesn’t seem to be completely random -- the highest-ranked song in one group was never the absolute bottom choice of another -- but high-ranked songs in one group could certainly end up near the bottom of another’s list. We can theorize that there is some quality threshold, but beyond that, social effects take over: Knowing that someone else likes something makes you more interested in it, and so some combination of early rankings and random variation among the groups creates a unique outcome in each social network. It’s our old friend path dependence in viral form.

Post-hoc, of course, we construct all sorts of reasons that popular things are popular. But as Watts points out, what we’re often doing is not so much explaining the popularity as describing the attributes of the popular thing: The Mona Lisa is popular because it’s so, well, Mona Lisa-esque. And those explanations tend to leave out the more random elements -- like the fact that the Mona Lisa wasn’t that popular until it was stolen in a famous museum robbery.

Maybe there’s no particular reason that millions of us wanted to spend our Thursday evening playing with a wedding-themed optical illusion, except that we all ended up doing just that. And most of us enjoyed it, so maybe randomness is not the right word. Maybe we should call it “serendipity” and leave it at that.

This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.

To contact the author on this story:
Megan McArdle at mmcardle3@bloomberg.net

To contact the editor on this story:
Brooke Sample at bsample1@bloomberg.net