How analyzing New Year's resolutions led me to the dark world of Twitter spambots
Originally posted on the SocialProvidence blog
As the end of 2014 approaches, social media is making its slow transition from the holidays to the new year. TV ads are transitioning from "this year's Kia holiday sales event" to "Gold's Gym; your first 6 weeks for just $19.99." BuzzFeed is replacing those "17 Cats Who Are Going To Eat All Of Santa's Cookies" listicles with "49 New Year's Resolutions Twentysomethings Should Consider." Target shelves previously lined with green and red candies are now filled with self-help dieting books and home gym machinery. That's right. It's time for New Year's Resolutions.
I'm not normally one to make or keep a New Year's Resolution (and apparently I'm not alone.) But I was curious about what most people are actually resolving to do. With so many gym membership ads, are people really resolving to lose weight? Or could these marketers be out of touch - are people actually resolving to travel more? I took a look at Twitter activity surrounding the phrase "New Year's resolution" to find out. Though I didn't find explicit answers to my questions, I took away some interesting key points:
1) More people declared their 2014 resolution on New Year's Eve than on New Year's Day
On December 31, 2013, more than 106,000 people talked about their New Year's resolution on Twitter, whereas on New Year's day, only about 87,000 people did. Maybe people are too busy living out their resolutions on January 1st, but if you want to be part of the 106,000 trendsetters this year, make sure to Tweet your new declaration by midnight Wednesday.
2) Lots of people want to stop eating chips
Using analysis of all tweets containing the phrase "New Year's resolution," I broke down what words were mentioned most often and produced the word cloud below. Among the camp of predictable words were "2014", "year", "better" and "change." Some surprisingly frequent mentions were "caribbean", "defeat", "pronounce", and "owlcity." In fact, 1.65% of tweets mentioned Owl City - that's over 3,000 people who mentioned Owl City in tandem with their New Year's resolution. And with 1.86% mentions, more than 3,500 people resolved to do something that involves chips (I'm going to assume it was to stop eating them.)
What's the best resolution you can make out of the high frequency words above? Mine is "This year I can't stop thinking: my New Year's Resolution will be to actually drink happy defeat in the caribbean for my life. Don't need chips"
3) There's a dark world of Twitter spambots
One of the first graphs I pulled from our New Year's resolution data was the one below. Naturally, frequency in mention of the phrase spikes between December 29th and January 2nd. Similarly predictable, mentions of resolutions slowly increase during the month of December and decrease during the month of January.
But what about those little spikes in activity throughout the year? What is causing people to mention New Year's resolutions on January 29th, February 5th, and May 22nd? Could there be some brilliant viral campaign that got people talking about their resolutions months later? So I did some investigating...
As it turns out, they're automated spambots. And there are tons of them.
It quickly became apparent that every spike in New Year's Resolution mentions during the "resolution offseason" (not in December or January) could be attributed to these spambots.
Each spike followed the same formula (or program). One spambot would tweet "My New Year's resolution" followed by a presumably spam link (I didn't want to check). Then thousands of spambots would retweet it. This would happen within one minute. I assume that Twitter would then sense the spambots and shut them down somewhere around the 3,000 retweets threshold, since most of the spikes only reached around 3,000 mentions.
I was not surprised to discover that spambots dominate these spikes in conversation around New Year's resolutions, but I was impressed that so many of these fake accounts do actually tweet, and they tweet as if they're teenagers. Someone has written a brilliant program that imitates your most insecure and angsty eighth grade away messages. Common phrases included "I love having this one special person who you know will always be here for you", "when boys hold your hand in public" and "Oh, you liked my picture from four months ago? You obviously weren't stalking me." My personal favorite was "Do u ever take a nice selfie on an iphone front camera but then it FLIPS the photo and u look like a gremlin". One fake profile even ventured into the land of great irony, lamenting an abundance of phonies in a world where she's trying to be true to herself:
As you look to make your 2015 resolution, keep in mind when to tweet (New Year's Eve), what to tweet about (some combination of the words "think", "caribbean", and "Owl City"), and make sure to keep your spambots from creating ironic caricatures of themselves ("so many fake people".)
Until next year, friends!
For another great spambot story, check out Alexis Madrigal's story in The Atlantic.