Will Facebook replace psychological testing and traditional surveys for social scientists researching the human condition?
Perhaps. Using computational methods, researchers can now tap into the huge amounts of information on the some 1.4 billion user profiles of Facebook, opening new avenues of study into human personality and behavior that can scale to proportions previously possible only with huge grants and tenure.
Although there is a wide range of social media environments, such as Twitter or YouTube, I have found that Facebook is particularly convenient for conducting social science research. Facebook profiles contain a wealth of demographic information and behavioral footprints that are, in most cases, much more personal than those in other environments. Also, the information found on Facebook profiles is of exceptionally high quality — invalid profile information is difficult to maintain in a network of friends who can challenge false assertions.
The breadth of Facebook data allows for discovering patterns that would be impossible to detect using traditional approaches. By analyzing Facebook Likes of millions of users, for example, we can reveal subtle patterns that would be difficult to identify using traditional surveys. For example, if someone likes The Matrix, a human analyst would have trouble translating it into a prediction of the given person’s character, but a computer model can put it into the context of millions of other users and would conclude, in this case, that the individual is likely to be intelligent and introverted.
The results, too, are often spot on. My research, published earlier this year in the Proceedings of the National Academy of Sciences, showed that a computer model based on Facebook Likes can predict personality traits better than one’s friends and family members. I further discuss how Facebook and other social media are changing research methods in “Facebook as a Research Tool for the Social Sciences,” published in American Psychologist in September.
Another advantage of Facebook-based research is the size of the population that we can tap into — currently 1.4 billion. That means that even those demographic groups that are underrepresented on Facebook, such as older or less educated, are still available in very large numbers. This population can be also accessed at a relatively low cost, especially if one includes viral features that encourage participants to invite their Facebook friends. In 2007, my colleague David Stillwell, a graduate student at the time, launched a Facebook app called myPersonality, which offered participants access to 25 psychological tests as well as feedback on their scores. He invited 150 of his Facebook friends at the time. Four years later, we had attracted an astounding 6 million participants. We make the database available to others and, so far, almost 200 researchers from more than 100 academic institutions have used data from it.
Yet using Facebook data has its flaws, too. First of all, while there is a lot of empirical evidence showing that, in general, Facebook profiles reflect the actual and not self-idealized image of their owners, researchers have to be aware that users can control and selectively remove information from their profiles, painting their lives as more (or less) idyllic than the reality. Second, the behavior of Facebook users is, to some extent, affected by the algorithms of the platform. For example, stories that appear on one’s news feed are clearly more likely to be liked. Thus, while studying Facebook data, researchers have to cautiously separate the effects of users’ preferences and behaviors from the effects of Facebook algorithms.
Perhaps the biggest issues with using Facebook for research are the ethical ones, and particularly protecting the privacy of its users. Of course, individuals must opt in to take a researcher’s Facebook-based personality test or to allow access to their Facebook profiles. However, when they hit that “I agree” button on the consent form, they may not realize just how much information they’re allowing the researchers to see. They may be fine with sharing their gender, location, and even political leanings, but they may not realize that a particular photo might reveal something about their health or sexual orientation.
That’s why researchers should shoulder the burden of being specific with prospective participants regarding what they could be sharing upon signing the consent form. The mere availability of data and participants’ willingness to share it does not grant researchers an automatic right to record and use it freely.
The privacy questions extend beyond the participants to their friends. Many elements of a Facebook profile, such as comments or photos, contain data about the user’s friends, who have not consented to participating in the research. Researchers may have access to such data, but should they use it? I believe that it is acceptable to use such data, as long as it is only used to learn about the participant and not the friend. For example, we should be able to note a friend’s gender to help us determine how many male and female friends the participant has. Or, if we see that friends have posted photos related to extreme sports, we might learn more about the participant’s interest in those activities. We’d focus not on the friends themselves, but rather on what they tell us about the participant. Certainly, we need to develop more guidelines pertaining to new privacy challenges posed by social media environments.
Despite the hurdles it presents, Facebook and other social media sites will only continue to grow as research tools. Compared with old-style laboratory-based research, Facebook provides a powerful approach to studying people. I am quite sure that one day, research based in a digital environment will become more widespread than traditional psychological experiments and studies.
Michal Kosinski is an assistant professor of organizational behavior at Stanford Graduate School of Business. “Facebook as a Research Tool for the Social Sciences” — written with Samuel D. Gosling of the University of Texas, and Sandra C. Matz, Vesselin Popov, and David Stillwell of the University of Cambridge — was published in the September issue of American Psychologist.