Monday 19 December 2011

Happiest days of 2011: Tweeting reveals our daily mood swings

If you wanted to measure daily rhythms in activity of mice, you could record wheel running behaviour. If you wanted to measure daily changes in hormone levels, you could take regular samples throughout the day and night. And these days apparently, if you want to measure daily mood swings in humans, you can look at what words they are tweeting.

Twitter is a 5-year-old social networking website, claiming over 100 million users who are tweeting an average of 230 million tweets a day. Scientists this year have taken advantage of this free public access to our real time moods and have used tweets to analyse our daily, weekly, and seasonal mood swings.

Cornell researchers Scott Golder and Michael Macy analysed 509 million messages tweeted between 2008-10 from 84 countries [1]. These tweets were published online by 2.4 million English speakers that had written between 25-400 messages.

They measured positive words (agree, fantastic, super) which showed tweeters delight/enthusiasm/activeness/alertness as well as negative associated words (afraid, mad, panic) which reflected writers' distress/fear/anger/guilt/disgust. Then they looked for changes in the amount of positive or negative words over the day and across seasons.

Their first finding was that positive words and negative words have different rhythms, showing that these moods are not different ends of the same spectrum, but more complex.

They found daily mood changes: people are more positive early in the morning. This happiness dips during the day and then rises again in the evening. This daily mood swing also happens at the weekend, albeit 2 hours later (after a lie-in?) suggesting that work isn't entirely to be blamed for our daily dip in enthusiasm. However, overall people are happier at weekends when they don't have to work and can sleep when they like.

Our daily mood swings as analysed from 509 million "tweets" from 2008-10. Adapted from Golder & Macy 2011
One of the real strengths in this analysis is they also looked at tweets from those living in the United Arab Emirates, where the weekend is on Friday and Saturday. They observed the same weekend trend as we have in the West, where the weekend is Saturday and Sunday.

Larks (morning risers) and owls (evening types) who tweet at different times of the day also have different rhythms in positive and negative tweets [1.1].

Golder & Macy's final findings show that when the days are getting longer, we get happier, but when the days are getting shorter our positivity decreases. However, there are no seasonal affects of negative words, so the "winter blues" seem to be a net result of less positive thoughts rather than more negative ones.

A more recently published study of the happiness of tweets over 3 years, has produced an incredibly detailed picture of rhythms of tweets [2].  The daily peak for 'coffee' tweets occurs between 8 and 9 am. Outliers of our daily happiness can also be picked up: these occur on holidays, such as Christmas, and during global tragedies, such as the tsunami.
"Coffee" is most frequently tweeted between 8 and 9am
Twitter and social media sites are becoming an invaluable resource to scientists studying our social rhythms. The data is generated real-time with no reliance on memory. They provide a larger data set, representing the general public better than small groups of American undergraduates. However, like any questionnaire, they do rely on the emotions that people publicly share, rather than what might be their innermost emotions.

Overall twitter is helping to reveal humanity's mood swings, and the universal effect sleep and the body clock have on our mood.

[1] “Diurnal and Seasonal Mood Vary with Work, Sleep, and Daylength Across Diverse Cultures.” Scott A. Golder and Michael W. Macy. Science, Vol. 333, Sept. 30, 2011

[2] "Temporal patterns of happiness and information in a global social network: Hedonometrics and Twitter" Dodds et al., PLoS ONE, Vol 6, e26752, 2011

No comments:

Post a Comment