Inside the Lab That’s Quantifying Happiness

0
3

In Mississippi, individuals tweet about cake and cookies an terrible lot; in Colorado, it’s noodles. In Mississippi, the most-tweeted exercise is consuming; in Colorado, it’s working, snowboarding, mountain climbing, snowboarding, and biking, in that order. In different phrases, the 2 states fall on reverse ends of the habits spectrum. If you had been to assign a caloric worth to each meals talked about in each tweet by the residents of the United States and a calories-burned worth to each exercise, after which totaled them up, you’d discover that Colorado tweets the very best caloric ratio within the nation and Mississippi the worst.

Sure, you’d be forgiven for doubting individuals’s honesty on Twitter. On these uncommon events once I destroy a complete pint of Ben and Jerry’s, I most assuredly don’t tweet about it. Likewise, I don’t attain for my cellphone each time I strap on a pair of skis.

And but there’s this: Mississippi has the worst rate of diabetes and heart disease within the nation and Colorado has the best. Mississippi has the second-highest percentage of obesity; Colorado has the bottom. Mississippi has the worst life expectancy within the nation; Colorado is close to the highest. Perhaps we’re being extra trustworthy on social media than we expect. And maybe social media has extra to inform us in regards to the state of the nation than we notice.

That’s the proposition of Peter Dodds and Chris Danforth, who co-direct the University of Vermont’s Computational Story Lab, a warren of whiteboards and grad college students in a good-looking brick constructing close to the shores of Lake Champlain. Dodds and Danforth are utilized mathematicians, however they might make a fairly good comedy duo. When I finished by the lab lately, each had been in working garments and cracking jokes. They have an abundance of curls between them and the wiry vitality of continual thinkers. They got here to UVM in 2006 to begin the Vermont Complex Systems Center, which crunches large numbers from large methods and appears for patterns. Out of that, they hatched the Computational Story Lab, which sifts by means of a few of that public information to discern the tales we’re telling ourselves. “It took us a while to come up with the name,” Dodds informed me as we shotgunned espresso and gazed into his MacBook. “We were going to be the Department of Recreational Truth.”

This 12 months, they teamed up with their PhD scholar Andy Reagan to launch the Lexicocalorimeter, an internet instrument that makes use of tweets to compute the energy in and energy out for each state. It’s no mere social gathering trick; the Story Labbers consider the Lexicocalorimeter has vital benefits over slower, extra conventional strategies of gathering well being information. “We don’t have to wait to look at statistics at the end of the year,” Danforth says. “This sort of data is available every day. We can tell if a public health campaign to invest in school nutrition is changing the way people talk about food or engage in activities.” For instance, what if Michael Bloomberg’s proposed ban on sodas bigger than 16 ounces had gone by means of in New York? Using conventional surveys and hospital studies, it will have taken years to measure the influence. But if the Lexicocalorimeter was tuned finely sufficient to precisely measure the adjustments in soda habits by neighborhood, then public well being officers may use it to focus on investments and modify the marketing campaign to cut back weight problems much more successfully.

Playing round with the Lexicocalorimeter is illuminating and sometimes horrifying—a glimpse of the unvarnished American character. Click on a state, and it shows the 200 phrases that made the largest distinction in that state’s calorie counts. In Louisiana (#48 in caloric steadiness), all people’s consuming chocolate, cookies, shrimp, and cake. Everybody’s consuming, interval. It’s one of many solely actions continuously talked about. In California (#12), they dance, run, hike, and bike, however they hardly ever sit or lie down. In my dwelling state of Vermont, the meals on the tip of all people’s thumbs is bacon, which might be an enormous a part of why we devour barely extra energy than the typical state. (In our protection, we additionally spend an inordinate period of time tweeting about beets, broccoli, and bananas.) Despite that, we’re pretty train obsessed, with—you guessed it—snowboarding main the best way. (We crush it on sledding, too.) All this provides us the third-best caloric ratio within the nation, behind Colorado and Wyoming. And certain sufficient, the well being numbers match: We have among the lowest charges of diabetes and weight problems and one of many highest life expectations.

In basic, all states are extra alike than we’d prefer to consider. “Watching TV or movie” is the most-tweeted exercise for each single state within the union, and “pizza” is the most-tweeted meals for each state besides Wyoming (cookies) and Mississippi (ice cream). Where a state’s particular person character actually shines is within the meals and actions talked about far kind of than common. Texas (#36) can’t cease tweeting about doughnuts; Maine (#5) is hooked on lobster. In actions, the mountain states do lots of working, the South is a strong block of consuming, New Jersey is all about “getting my nails done,” and Delaware distinguishes itself with “talking on the phone.”

Dodds and Danforth acknowledge their strategies are usually not excellent. The butter on the lobster doesn’t get counted. There’s no approach of calculating if someone ran 1 mile or ten. But while you’re speaking tens of hundreds of thousands of tweets per day over the complete vary of demographics, the inaccuracies even out—at the least as a lot as they do in comparison with the opposite, equally flawed methods of measuring society’s consuming and exercising habits. As Dodds factors out, the numbers communicate for themselves: The Lexicocalorimeter correlates extraordinarily properly with charges of diabetes and weight problems. “The ridiculous thing about this,” he says, “is that it works.”


We stay in unusual instances. “People leave so much of their id on the web,” Danforth marveled to me, “and they share it openly. That’s enabled a whole host of new instruments to try to understand what’s predictable about our behavior. And it turns out a lot is. As much as we think we’re really complex, people have very structured ways of behaving. The way we move around the earth is very predictable. The way we use language is very predictable.”

For instance, Barack Obama’s approval scores over his presidency strongly correlate with the sentiment of tweets about him 3 months upfront of the approval polls. In different phrases, in case you’d been a savvy politico with a instrument for measuring tweets, you’d have had precious intel months forward of anybody else. “It’s an amazing time in social science because of the data available,” Dodds says. “It’s opened up a window that we absolutely did not have access to before.”

twitter-happiness-measure.jpg
(Courtesy of the Computational Story Lab)

That’s the thought behind the UVM staff’s Hedonometer, which surveys the nation’s tweets every day and calculates a happiness rating for every. The staff had individuals charge 10,000 phrases on a happiness scale of 1 (unhappy) to 9 (joyful). Most phrases are impartial. The happiest words are “laughter” (8.50), “happiness” (8.44), and “love” (8.42). “Hahaha” will get a 7.94, placing it a bit increased than “kisses” (7.74). The largest negatives are “terrorist” (1.30), “suicide” (1.30), and “rape” (1.44). “Shit” will get a 2.50, “bitch” a 3.14, and “fuck” a surprisingly respectable 4.14. Fuck yeah! “Swearing is really important,” Dodds says.

All this provides as much as a mesmerizing sine wave monitoring the nation’s temper from 2009 (the fledging of Twitter) to the current. “One of our goals was to provide a snapshot of the public’s response to something,” Danforth defined, “the texture of the day.” Most common days fall right into a slender band with a mean happiness stage round six, although Saturdays are persistently the happiest days of the week and Tuesdays the grumpiest. There’s additionally a day by day sample, with happiness ranges hovering round 5 and 6 a.m., once we’re all newly optimistic in regards to the day, after which plunging all through the morning and night as actuality units in, reaching a trough of despair round 11 p.m. “The wheels kind of come off,” says Dodds. “We call it the daily unraveling of the human mind.”

You can even parse this by state. The happiest state is—unsurprisingly—Hawaii. The backside dwellers are, as soon as once more, Mississippi and Louisiana, although Delaware will get a stunning bronze for melancholy. The West is joyful, each coastal and mountains, whereas the South and Midwest are sad. Only Tennessee bucks the pattern, an island of smiles in a sea of Southern gloom.

The happiest day of the 12 months is all the time Christmas, when the Hedonometer spikes as phrases like “Christmas,” “happy,” “family,” and “love” flood the ether. The 5 unhappiest days since 2009: the capturing at Sandy Hook Elementary School, the Boston Marathon bombings, the Orlando nightclub assault, the capturing of Dallas law enforcement officials, and the election of Donald Trump.

National happiness will not be constant. We had been fairly joyful from 2009 to 2011, regardless of the Great Recession. Then our temper darkened from 2011 to 2014, however we got here out of it: The Hedonometer surged! The 12 months 2015 was a comparatively joyful 1, and the nice emotions stored getting into 2016—till the election took over. Since then, indicators have been rising that one thing horrible is occurring to the American psyche. We’ve by no means been so erratic, with the usually easy blips of the Hedonometer beginning to twitch like somebody failing a lie-detector take a look at. And as of this writing, we’re sinking into an unprecedented malaise.

That is, in case you consider the Hedonometer. On the face of it, measuring one thing as intangible as happiness sounds absurd. Yet, as with the Lexicocalorimeter, the Hedonometer matches “real world” measures such because the Gallup Well-Being Index (which polls individuals on issues like life satisfaction and private well being) and the Peace Index (which surveys charges of homicides, violent crime, and incarceration).

History is stuffed with ideas—from longitude to time—that appeared imprecise till the suitable instrument got here alongside. Even temperature, which to us appears goal, was thought-about unmeasurable for hundreds of years. “People thought you couldn’t do it,” Dodds says. “Because it’s too multifaceted, and the first thermometers were awful.” But finally our devices improved.

Dodds and Danforth see no cause why happiness can’t even be quantifiable. “We’re carrying around these phones that are sensing so much of our behavior,” says Danforth. “Tone of voice. Who we talk to. The types of words we use. We’re trying to push on a few areas and see what’s predictable, both on the population scale and for individuals.” And what they’re discovering is that our telephones have turn out to be surprisingly good devices for taking our emotional temperatures. “Can we tell you’re about to experience an episode of depression based on your social media behavior? Maybe your friends can’t see it, maybe you don’t even realize it, but you’ve started to communicate with a smaller group socially, or you’re not moving around the earth as much.”

By analyzing the tweets of each depressed and wholesome people, the Story Lab has developed algorithms that may precisely establish melancholy months earlier than precise diagnoses by psychological well being practitioners. They’ve even achieved it with Instagram, discovering that depressed people usually tend to put up pictures which can be bluer, grayer, and darker. Their technique outperformed skilled practitioners at figuring out beforehand undiagnosed melancholy. The lab is now partnering with a psychiatrist at UVM who hopes to make use of the algorithm to go looking the social media historical past of ER guests (who give their consent) to foretell suicidal habits.


One of the clearest indicators that the Hedonometer is on to one thing is how properly it really works with media apart from Twitter. The Story Lab has analyzed the phrases in 10,000 books and 1,000 film scripts, and it precisely kinds the feel-goods from the nihilists. Sex and the City is the cheeriest film of all, powered by phrases like love, smiles, wedding ceremony, lovely, and, sure, intercourse. At the underside of the listing now we have grim fare like Commando, Day of the Dead, The Bourne Ultimatum, and Omega Man.

Yet for grimness, none of these can contact Outside’s masterpiece of masochism, “Bury My Pride at Wounded Knees,” by Mark Jenkins, about competing within the 2010 Death Race. I requested Reagan, Dodds, and Danforth to take the emotional temperature of 49 classic Outside stories, which run the gamut from Steve Rinella’s joyful paean to Argentinian steak (“Me, Myself, and Ribeye”) to Jenkins’ mudfest, which begins “I unintentionally pitchfork a clod of manure into my mouth” and goes downhill from there.

into-thin-air-happiness-measure.jpg
(Courtesy of the Computational Story Lab)

Beyond a bit’s general happiness rating, the Hedonometer permits us to chart its emotional journey. Here’s “Into Thin Air,” maybe the last word Outside basic, which begins off joyful (Everest journey!), plunges a couple of 1/3 of the best way by means of the story as Krakauer reaches the perilous Lhotse Face (“It was here that we had our first encounter with death on the mountain”), soars on the midway level (summit!), after which tanks far and quick as storms transfer in, errors are made, and other people die.

Not solely does the fall-rise-fall emotional arc of “Into Thin Air” properly mirror the Himalayas, nevertheless it additionally occurs to be a great instance of a basic narrative arc. Riffing off Kurt Vonnegut’s famous talk on the shapes of archetypal tales, the Story Labbers got here up with six arcs that tales are inclined to observe: Rags-to-Riches (rise), Tragedy (fall), Man-in-a-Hole (fall-rise), Icarus (rise-fall), Cinderella (rise-fall-rise), and Oedipus (fall-rise-fall). Encouragingly, an evaluation of the bestseller lists discovered that the extra complicated narratives (Cinderella and Oedipus) are inclined to promote higher than the less complicated ones.

This fall, the UVM staff might be engaged on teasing much more tales out of the information we share: Can monetary crashes be predicted forward of time? When does pretend information trump actual information? How does a society choose the story it tells about itself? The questions are removed from trivial. “Humans are storytelling organisms,” Dodds says. It’s how we study who we’re. As the Story Lab will get even higher at discovering the alerts in our noise, let’s hope we like what we uncover.

(Editor references)

Leave a Reply