Marketers may struggle to use data to understand what consumers want to buy, but a computer scientist told the CMDC Vision 20/20 conference in Toronto on Tuesday about a much more ambitious undertaking: a study of social media to see whether or not recovering alcoholics will stay sober.
Though the full results have yet to be published, the University of Maryland’s Dr. Jennifer Golbeck told the audience of media directors and agencies about research that tracked what approximately 300 people said on Twitter prior to attending their first meeting of Alcoholics Anonymous (AA). This was then correlated with whether or not the individuals in question managed to avoid drinking for 90 days, or even an entire year.
“I realize this takes the whole ‘anonymous’ element out of it,” she joked, adding that while the data is still being assessed it’s an example of how such data can provide real value. “The big takeaways is that these algorithms are not bad. They can be used in all kinds of important ways.”
This includes targeted advertising, of course which Golbeck cited as achieving major progress in her personal demographic. “I now see ads for shoes, which is much better than, say, fertility treatments, which is I guess what companies thought women in their 30s were supposed to see when they went online.”
On the other hand Golbeck, who is director of her school’s Human-Computer Ineraction Lab, admitted that some of the information that comes through studying social media behaviour isn’t always clear-cut. She cited a third-party study that tried to identify those with high intelligence from those with low intelligence by looking at pages they “like” on Facebook. The smart people tended to “like” pages such as Science, Thunderstorms and . . Curly Fries. The No. 1 Facebook page associated with low intelligence, meanwhile, was aimed at mothers who enjoy being a parent.
“There can be a statistical connection, but not a logical one,” Golbeck said.
The real art to sifting through the data may involve figuring out what activities suggest a particular attitude or emotion. Another Twitter study Golbeck talked about was trying to discover which new mothers might experience post-partum depression, for instance.
“For the women who experienced post-partum depression, the number of questions they asked (on Twitter) went up,” Golbeck noted. “For those who didn’t experience post-partum depression, there were more likes, more retweets.”
The key thing for brands, Golbeck recommended, is ensuring the data that gets collected and analyzed doesn’t “cross the line” in terms of consumers’ preferences around privacy. This is largely about being clear about what marketers are tracking. Golbeck said she has personally used apps on Facebook that override her strictest privacy settings.
“If anyone should know how to make sure this doesn’t happen, it’s me,” she said. “I work on this full time.”
CMDC’s Vision 20/20 conference with a theme of “The Speed of Culture,” wrapped up Tuesday afternoon.