Measuring sentiment has always been tricky for scientists. There’s no single metric to indicate whether someone will be happy or sad, and people’s desires can differ so dramatically that even something like, say, winning a million dollars, might not translate into actual happiness. The country of Bhutan has tried to aggregate its people’s happiness into a single number—based on the economy, environmental wellness and political stability—known as Gross National Happiness. Yet the model isn’t exactly conclusive, nor easily transferable to other countries.
Social media might be the answer. A new study from the University of Vermont realized that analyzing tweets might offer clues about aggregate happiness in a given U.S. state. The results are bemusing. America’s happiest state is, unsurprisingly, Hawaii, followed by Maine, Nevada, Utah and Vermont. Bringing up the caboose as the saddest states are Delaware, Maryland, Mississippi, and, dead last, Louisiana.
The researchers geo-tagged 10 million tweets from 373 urban areas in America during 2011. Then they gave each word a value of positiveness on a scale from one (mopey) to nine (elated). ‘Rainbow,’ for instance, had the happiest score at 8.1. ‘Earthquake’ was the lowest at 1.8. Words like “the” or “thereof” without inherent sentiment score somewhere in the middle. For statistical buffs, the calculation is complex. You can find it in the study, here.
Sentiment measuring isn’t terribly new. Odewire, a project of Intelligent Optimist Magazine, developed a news screening search engine for people who only want good news. The analytics company General Sentiment has refined software for measuring favorability mentions of TV shows, sports stars and political candidates.
The begging question: is it actually accurate? Algorithms can do a lot of things these days like forecast your travel preferences and automate your Christmas gift list. But measuring the happiness of specific words takes a leap that seems to ignore context. Anyone can use the word rainbow, but does that mean they’re happy (“This weather is terrible, I wish I could see a rainbow,” might, for instance be considered pretty joyful; “Barack Obama hates gun violence,” packs a lot of negative-sounding words). Then there’s the drawback of just testing social media, users of which skew younger and more susceptible to self doubt, insecurity and melodrama. Getting a computer to give a mosaic can be helpful and interesting. Still, in a world full of nuance, context is often something a computer—and sometimes even a human—can simply find much harder to pick up.