An individual’s Facebook “Likes” can be used to quickly and easily create an entire demographic profile, but at what cost?
Clicking the “Like” button on Facebook reveals a lot more about you than just a personal preference. In fact, according to a new study led by researchers at the University of Cambridge, Facebook “Likes” can be used to predict everything from your sexual orientation to your religious and political views. And while this information can be used to help improve products, services, and online marketing, it also poses a serious threat to personal privacy and data ownership.
According to researchers, Facebook “Likes” are a way for users to express positive feelings about online content, such as photos, friends’ status updates, and Facebook pages of products, sports, musicians, books, restaurants, or popular websites.
“This study demonstrates the degree to which relatively basic digital records of human behavior can be used to automatically and accurately estimate a wide range of personal attributes that people would typically assume to be private,” the study authors wrote.
More than 58,000 volunteers provided their Facebook “Likes,” detailed demographic profiles, and the results of several psychometric tests. Researchers were able to use that data to accurately predict many of their personal traits and preferences, including ethnicity, intelligence, happiness, use of addictive substances, parental separation, age, and gender.
Study co-author Michal Kosinski said that while several other studies have examined the relationship between online behavior and psychological traits and demographics, this is the first study to look at such a wide variety of traits, based on such a large sample.
“My favorite part is that, whereas previous researchers have linked behavior online with personal traits, Facebook ‘Likes’ have a meaning that we can use to understand the psychology behind what people do,” Kosinski said in an interview with Healthline.
For example, the team found that in the supplemental table for ‘parents separated at 21,’ some of the most predictive ‘Likes’ of parental separation are liking statements such as, “I’m sorry I love you” and, “If I’m with you then I’m with you I don’t want anybody else.”
“Although our prediction is not very good—60 percent, which is just above chance at 50 percent—it gives us a poignant insight into the effects that parental breakup has on children even after they grow up,” he said. “It was surprising to us that parental breakup has any effect at all on the things you choose to ‘Like.’ It suggests that online behavior can be a goldmine for social scientists, improving our understanding of humans.”
However, this study also demonstrates that we may be sharing more with others than we’d like, revealing highly personal traits through seemingly innocent online behavior, said Kosinski.
“Governments, companies, or even other individuals (e.g. your Facebook friends or Twitter followers) can use a simple software to reveal the personal and highly sensitive traits of just about anyone,” he said. “Imagine governments inferring citizens political views (~85% accuracy), religious governments inferring religion (~82% accuracy) or sexual orientation (~88% accuracy), and so on. The sheer possibility of this happening can deter many people from using digital technologies, ruin the trust between individuals and institutions, and thus hamper technological progress and the economy.”
Using demographic information from the volunteers’ Facebook profiles and other traits, such as intelligence, personality, and satisfaction with life that were measured in online surveys and tests, Kosinski and his colleagues “accurately predicted study participants’ gender, ethnic origin, and sexual orientation, correctly identifying males and females in 93 percent of the cases, African Americans and Caucasians in 95 percent of the cases, and homosexual and heterosexual men in 88 percent of the cases. The model also correctly classified Democrats and Republicans as well as Christians and Muslims in more than 80 percent of the cases, but was less accurate at predicting relationship status, substance abuse, and parents’ relationship status.”
In the future, these findings could be used to create a detailed profile of each individual accessing a given website or media channel for targeted marketing, or to screen millions of candidates at once before inviting one in for an interview. Such personal information canould even change the way humans interact with computers, yielding websites and machines that can detect a user’s personality and adjust their behavior accordingly, said Kosinski.
“Imagine a new smartphone automatically adjusting its settings to the user based on the predicted personality profile. Or a new car tuning the settings of its engine to best suit the personality of the driver,” he said. “Obviously, similar technologies are already being used in the industry. For example the recommender systems—books suggestions on Amazon, or one selecting the most interesting status updates to put on your Facebook newsfeed. Recommender systems work by trying to figure out who you are and what you want based on your previous behavior.”
While this technology might make some of us wary about what we share or ‘Like’ online, Kosinski believes that being able to predict individual traits offers advantages that greatly outweigh the risks, and that this research doesn’t mean we will lose all control of our personal data.
“I hope that these results will help companies and policy makers to shape technology in a way that gives individuals a full control over what information they reveal,” he said.