All those postings, tweets, Flickr uploads and other social media interactions we have every day add up to a wealth of opportunity for scientists. Simply by inhabiting our digital haunts, observers can learn volumes about how we feel, determine where it’s safe to eat, even plot the passing of a storm.
Often, gathering the intelligence doesn’t even require an app. Recent innovations range from deliberate crowdsourcing to sweeping but incidental gathering and plotting. Here are some of the approaches reported in recent studies.
Tweets for safe eats: Scientists at the University of Rochester found that by analyzing tweets from New York City diners, they could figure out how likely you were to contract food poisoning if you patronized a particular restaurant.
Over a four-month period, they collected 3.8 million tweets from more than 94,000 unique users, traced 23,000 restaurant visitors and found 480 reports of likely food poisoning.
They also noted the tweeting matched somewhat well with public inspections reported by the health department — about a third of the inspection failures could be reliably predicted from Twitter data. The remaining scores show some disagreement, but the researchers argue that the tweets may have detected foodborne illness that restaurant inspectors might not.
The research, being presented this week at the Conference on Human Computation and Crowdsourcing in Palm Springs, suggests tweets might offer another approach to food-safety monitoring.
Texting from your happy spot: Researchers at Princeton University designed a multinational project using mobile phones to measure happiness. They used geo-location to track 270 volunteers in 13 nations. They designed an app that, over three weeks, periodically asked: “How happy are you?” with answers on a scale of zero to 5.
The idea, researchers said, was not so much about defining the source of happiness as proving a concept: that mobile phones might better capture feelings “in the moment” than phone or in-person surveys. The team found that men tended to describe themselves as less happy when they were farther from home. Distance from home didn’t seem to matter to women in the study, reported in the journal Demography in June.
What Facebook phrasing says about users: Words and phrases such as “party,” “great night” and “hit me up” were among the expressions that marked extroverts, University of Pennsylvania researchers found in analyzing language used by 75,000 Facebook users who’d completed a personality questionnaire. Researchers’ computer models used word clouds or clusters to predict age, sex and, to some extent, personality profiles. They correctly predicted gender 92 percent of the time; they could estimate ages within three years of the correct number more than half of the time. Their findings appeared in the journal PLOS One in September.
Photo-sharing up a storm: Another recent study tapped the photo-sharing site Flickr track the passage of Hurricane Sandy in late October 2012.
Research published in Scientific Reports Nov. 5 showed a strong connection between images related to Hurricane Sandy posted on Flickr and the dropping atmospheric pressure that marked the superstorm’s arrival over New Jersey last year.
Researchers at the Warwick Business School in Britain and several other institutions looked at all photos tagged “Hurricane Sandy” “hurricane” or “Sandy” from Oct. 20-Nov. 20, 2012. They found the highest number of pictures were posted in the same hour that Sandy made landfall (Oct. 29) and pressure dropped. Photos tailed off as the storm moved inland and atmospheric pressure rose.
Researchers said the tight correlation between postings and air pressure could be useful in tracking the path and intensity of a storm, particularly where instruments are scarce.
Photos posted on Facebook were used by University of Georgia researchers to study how debris from a Southern tornado outbreak in 2011 was blown hundreds of miles away. The study involved nearly 1,000 photos, documents and other items from destroyed homes catalogued and posted on the website by an Alabama volunteer and later identified by family members.
Gaining understanding about how tornados move debris could help scientists improve warnings and plot risks if a storm struck a site that contains hazardous material, for instance. The Bulletin of the American Meteorological Society published the report last spring.