Verified by 
For a writer, it can be frustrating to slave over a story and then watch it flop online, slipping into a site's bottomless archives, practically unread. And watching something you write go viral is always a kick, even if it's an off-hand blog post. So it's hard not to keep an eye toward social contagion when conceiving a piece. And for publishers, noting which articles garner page views (and ad revenue) can mean sink or swim. Fortunately there's a new piece of research to guide us in taking over the Web.
Jonah Berger and Katherine Milkman of Wharton analyzed all 7,500 articles that appeared on the New York Times home page from 8/30/08 to 2/15/09 and observed which qualities put articles on the most-emailed list.
They measured positivity and emotionality using an algorithm that counted positive, negative, and emotional words in all the articles. They also recorded positioning on the homepage and time spent there. For three thousand articles, humans read summaries and rated the stories on how practical, surprising, and awe-inspiring they were. From their paper:
Examples of highly awe-inspiring articles include titles such as: “From Old Vials, New Hints on Origin of Life” and “Now in Sight: Far-Off Planets”. Examples of articles that scored highly on practicality include: “Vacation Plans That Avoid the Pump” and “2008 Holiday Gift Guide”. Examples of articles rated as extremely surprising include: “Iverson Adds Scoring and Who-Knows-What to Pistons” and “Day Care May Offer Asthma Protection.” Examples of highly positive articles include: “Wide-Eyed New Arrivals Falling in Love With the City” and “Free Films on Imdb,” while examples of extremely negative articles include: “Web Rumors Tied to Korean Actress's Suicide” and “Stampede in India Kills at Least 147.” Examples of highly emotional articles include titles such as: “Redefining Depression as Mere Sadness” and “Brad Pitt Supports Same-Sex Marriage.”
Berger and Milkman found that the biggest contributor to going viral was the ability to inspire awe--a sense of vastness that challenges one's world view. (See below* for how they explained "awe-inspiring" to participants.) Take a look at the included graph. To create this chart (as Milkman explained it to me), they took a hypothetical article with a 15% chance of making the most-emailed list and used their data to model what the effect on virality would be if it were more or less positive, say, or surprising. You can see that positivity, usefulness, shock-value, emotionality, and placement at the top of the homepage each has a moderate effect. But look at the rogue gray line. Being two standard deviations above the mean on ability to inspire awe--i.e. in the top 2.3% of articles on that measure--nearly doubles an article's chances of making the most-emailed list.
In a previous post I reported that awe can enhance a sense of social connection, so it makes sense that people would want to share awe-inspiring articles with each other, which then perpetuates the cycle.
To publications considering cutting science coverage, consider this: "Science articles were disproportionately likely to make the Times' most emailed list (30% vs. a baseline rate of 20%) during the period studied, and articles about things like far-off planets and emotion in animals were particularly viral."
I think the path to saving the media industry is clear: Clone Dennis Overbye.
(Just stay away from science fiction.)
*Awe-Inspiring
Articles vary in how much they inspire awe. Such articles teach people things or share a perspective that opens people's minds. This knowledge may not change how people live their everyday life, and thus may have little practical value, but it is more about knowledge for knowledge's sake. For example, reading an article about how the human mind works, or the structure of the universe is unlikely to change what people eat or what they do this weekend, but it will open up their mind to new possibilities and insights. Articles that are high on this dimension may evoke feelings of awe and wonder, and even articles that are mildly high on this dimension may just shift how the readers sees things, even if it doesn't really directly have practical utility.
[Updated to include Milkman's explanation of the model.]
