Twitter is now so big, and so constantly used, that two Cornell University researchers were able to use it as a sort of “global mood ring” to monitor the world’s feelings.
By analyzing the tweets of 2.4 million people in 84 countries, they report, they found that people generally wake up in good spirits, but things go downhill as the workday goes on. On weekends the pattern holds as well, though everything happens two hours later because people sleep in.
The patterns were consistent across the globe, they say, despite widely varying cultures and religions.
The researchers, graduate student Scott Golder and sociology professor Michael Macy, say they ran 509 million tweets through a computer program designed to discern moods from the people’s use of key words. The results are published in this week’s edition of the journal Science.
“People criticize the Internet for being mundane or filled with gossip, but that’s really not so,” said Golder in a telephone interview. “The Internet records everything, so Twitter is a giant archive of time-coded conversations.”
Golder set up a website — http://timeu.se – and invites you to try it out for yourself. Try entering a word, such as “happy” or “breakfast” or “commute,” and see what time of day it was used most often. Click “submit query” and you should get a graph like this:
A quick analysis: People tweeted the word “happy” most often around 7 a.m. or 8 a.m., and then its use tailed off. Are they bummed by the trip to work?
Try other words. (The word “sex,” perhaps predictably, craters during daytime hours, and rises overnight.)
Some patterns Golder teased out:
- People report they eat bacon more than sausage at breakfast, and Cheerios more often than Frosted Flakes.
- If people tweet the word “beer,” there’s a seven-hour lag, on average, before there’s a spike in the word “drunk.”
- Commuting on Friday isn’t so bad — but the trip home stinks.
Of course, many tweets are gibberish, or trivial, or marketing messages sent by corporate users. Lady Gaga and Charlie Sheen use it to talk to their fans; President Obama (or the staffers tweeting on his behalf) use it to attract voters.
But the researchers said there are so many tweets that there were more than enough to show mood patterns around the world. They confirmed the weekend mood boost, for instance, by looking at traffic from the United Arab Emirates, where weekends are celebrated on Friday and Saturday instead of Saturday and Sunday.
Golder readily conceded that the survey had its limits: “We’re measuring the expression of something, not the action itself.”
He added that he can’t do anything about people like, say, me (I was a bit too rushed this morning to tweet whether I was cheerful or grumpy or had a headache). But Twitter turns out to be a massive, if chaotic, database of people’s feelings, augmenting studies in which researchers ask people to recall how they felt.
“It’s a giant conversation,” he said. “Most of the conversations we have in real life are routine or mundane, but we don’t record them in real time on the Internet.”