The Affordable Care Act has spurred a frenzy of Internet activity over the past few years as partisans on both rabble-roused about the now-infamous technical failures of Healthcare.gov.
With the dust still settling, Facebook scientists released a paper on January 8 showing what made Obamacare memes spread. Namely: To get a meme to spread among liberals, zombies and sci-fi stuff is what primes them best. As for conservatives: They're more likely to hit the “Share” button if your meme includes alcohol.
In the publicly available paper, Facebook data scientists Lada Adamic, Thomas Lento, Eytan Adar, and Pauline Ng anonymized versions of 1.4 million Facebook status updates and analyzed them to see how their memes mutated and spread around the social network. The results show identifiable patterns in the way memes and status updates change as they spread from user to user. In the paper, the four researchers explicitly compare it to the way genes evolve through mutations.
Approximately 470,000 Facebook users posted the following status update in September, 2009:
“No one should die because they cannot afford health care, and no one should go broke because they get sick. If you agree, post this as your status for the rest of the day.”
Using anonymized status updates pulled from Facebook's user traffic--basically, status update archives that had a unique identifier but that couldn't be traced back to a specific Facebook user---the team found 121,605 variant versions of that string of text which appeared in 1.14 million status updates.
The most popular variants inserted someone's name. For example, "Sam thinks that no one should die." Others made small linguistic changes, such as changing "the rest of the day" to "the next 24 hours," a variation Facebook's data team believes was made by a user in the late evening hours.
By tracking when the variant status updates were made and what the connections were between the people posting them, Facebook created a visualization that resembles a phylogenetic tree. In the chart shown below, blue variants are strings of text with "the rest of the day," red variants are "the next 24 hours," and purple variants have other wording altogether.
Then Facebook's data team discovered that small turns of phrase led to users posting memes to their status feeds much more often. As the table of 4-grams below shows, certain phrases, such as the use of the words “your status,” led to users sharing these updates at a higher rate. But even more intriguingly--and not displayed on the chart--a negative correlation was found between mild spelling errors in a meme and its popularity. The more misspellings a variant had, the less likely it was to be shared by Facebook friends.
Even more interesting--though subjective--was the fact that variants which mentioned certain themes like booze or zombies skewed in one direction or another politically. As the chart below shows, science fiction-themed variants skewed liberal (blue), and alcohol-themed variants skewed conservative (red).
[Image: Flickr user Lars Plougmann]