(Screenshot by Michelle Starr/CNET Australia)
A new website pulls keyword data from Twitter to measure happiness around the globe.
There are many measurable quantities in the universe, but we would not have called "happiness" one of these. However, since 2008, a team of scientists from the University of Vermont and The Mitre Corporation, led by mathematicians Chris Danforth and Peter Dodds, have been figuring out a way to try. Together, they've been working on a piece of software called the Hedonometer that measures — and graphs — data pulled from Twitter Garden Hose, a random sample of 10 per cent of all tweets, to gauge how happy the world is on any given day.
A paid team of volunteers from Amazon's Mechanical Turk service analysed 10,000 of the most-used English words on Twitter, assigning a numerical value to the happiness level of key emotional words, such as "sad", "party", "scary" and "win". Each day, the tool crawls through the random tweets — some 100 million words — for these key terms, totting up the overall score. And it seems to work — the day of the Boston Marathon bombing, for example, was the saddest day in the nearly five years the team has been collating data.
Of course, there are some pretty serious limitations. Not everyone, especially in developing nations, has access to Twitter; and even if they did, the app is confined to English-writing users only, leaving out vast swathes of the global population. It would be safer to say that the app measures the happiness of relatively affluent, predominantly English-speaking people — and, it seems, predominantly in the US.
This is evidenced in the graphs. Christmas is routinely the happiest day of the year, with New Year's Eve and Thanksgiving about on par for second. There are also spikes for Valentine's Day, Independence Day and the US Mother's and Father's Days.
Nor can it take into account contextual cues. The word "love", for example, can be meant both sincerely and sarcastically.
However, the Hedonometer team is currently analysing and scoring the most frequently used words in a dozen other languages, with hopes to add them to the app by the end of this year — and it is also working on adding two-word expressions. It will be adding more data streams to the feed, such as Google Trends, Bitly, the New York Times and CNN transcripts. However, it would be good to see a bit more of a global focus on a tool that claims to be about the world.
Even within its limitations, though, the tool is interesting to look at. That Thursday and Monday are more or less similar in terms of unhappiness level rings true when you think about it, but it would not have occurred to us to consider it. Friday and Saturday are also about on par, which indicates that the biggest factor contributing to unhappiness within the data set is work.
It's also interesting to note that the graph seems to be on an overall downward trend since 2009; we are guessing that the relative low on which it started was due to tweaking the keywords and data collation. And although Danforth and Dodds note that the tool could be used by reporters, policymakers and academics, we can see the most intriguing value in the latter of the three.
We can also already see some improvements that could be made to the tool. Location tools, so that you can sort data by region. An averaging tool that lets you see the average happiness level of a week, month or year — and the ability to compare those side-by-side. A list of the most commonly used words on both ends of the scale. It's early days yet, and even Danforth and Dodds admit that the Hedonometer has a lot of potential to be explored.
"We're not trying to tell you that contentment is better than happiness — we're not trying to define the word," said Danforth. "We're just saying we're measuring something important and interesting. And now, sharing it with the world."