If somebody is having a "case of the Mondays" these days it's pretty certain you'll know about it in 140 characters or less. While this may seem like a no-brainer, scientists set out to quantify what we have all intuitively guessed — Twitter shows recognizable patterns on how we feel throughout the day.
Sociologists at Cornell University, reporting in the journal Science, have identified that people's daily moods seem to cross both cultural and environmental boundaries. Unique too, is this study is the first to quantify users' moods using text-only data.
Everyone knows there is a mountain of data being collected and studied about our every move on the interwebs. This study, headed by Scott A. Golder and Michael W. Macy, analyzed tweets from over 2 million average users in 84 countries. They collected over 400 tweets, written in English, between February 2008 and January 2010. What made this study unique from previous ones is that it looked at universal data where previous studies had looked at broad periods of time or in limited slices of data — such as in just one region or day part.
It's no big surprise people's tweets expressed a lower mood level at the beginning of the working week and grew more positive over the weekend. People expressed their happiest thoughts at the beginning of the day between 6:00 to 9:00 A.M. (we are wondering if that first cup of coffee figures in somewhere) and drifted to a slump in the afternoon around 3:00 to 4:00 P.M. and then rising again after dinner. It also confirmed a mood change with the seasons likely due to negative feelings towards the shortening of the days.
Twitter allowed access to the tweets and in a marriage of scientists and software developers they created a process of analyzing both emotional words such as "happy" or "bummed" and also factored emoticons into the equation. After weeding out the small populations of "early-birds" and "night-owls," the scientists were able to begin to aggregate the trends.
They also took great care to account for shifts in activity over the weekends (tweets activity occurring slightly later in the morning) and to adjust for countries where the weekends didn't fall on Saturday or Sunday. While previous studies have shown similar trends, it is these careful nuances of data analysis that have drawn attention to this report.
The Internet is a fickle mistress — often highlighting fleeting moments or news/information the Tweeter thinks people want to hear. Social media observers and probably our own common sense should tell us to take things we see and hear with a grain of salt. As much as we can slice and dice the data we can't extrapolate everything the user intended. Sarcasm, wit and sheer stupidity often can't be measured just by looking at words or waves on a chart.