Facebook's getting sketchier by the day.
Their methods seem simple enough: combine U.S. Census data with a list of users' first and last names. The team narrowed its analysis down to the country's four largest ethnic groups: whites, blacks, Asian/Pacific Islanders, and Latinos. With the help of the census records, they found, for instance, that someone with the last name "Mueller" has a 97 percent chance of being white, while a person with the last name "Washington" has over an 89 percent chance of being black.
First names make the search even more accurate. The name LaToya, for instance? Probably black.
The findings were instrumental in predicting users' friend networks, which all boils down to more advertising revenue for the company. Researchers found that users are more diverse than a few years ago, but are more likely to be friends with people of the same ethnicity. Same goes for romantic relationships. So, yeah, more proof that online social life mirrors that of the offline world.
Meanwhile, certain ethnic groups are more likely to behave in specific ways while logged in. Asian/Pacific Islanders, for instance, were said to engage in "unexpectedly high number of wall, video, note, gift, comment and group-sharing actions."
All this boils down to marketing, of course. Predicting people's behaviors is a massive part of Facebook's business plan, and the next step is reportedly to allow marketers to target users based on ethnicity, instead of just age, location, and known preferences.
The team of researches recently presented their findings at a social media conference organized by the Association for the Advancement of Artificial Intelligence in Washington, D.C., where they won the award for best paper.
So what do you think? Is it too creepy or all in the name of serving you better?