Using Social Media to Measure Personality: A New Chapter of Old Fiction



On February 19th 2015, CBS Evening News presented a story about using social media to measure personality. It featured a Stanford professor using algorithms on over 86,000 Facebook users to measure their personalities based on what they ‘Like’. Those who click on Shakespeare and 2001: A Space Odyssey were described as artistic, whereas those who clicked on Rush Limbaugh and Ford were described as conventional. Liking boxing was linked to being organized, and liking vampires was an indicator of being spontaneous. The story also featured similar work at the University of Cambridge, and a New York-based consulting firm using social media to screen applicants. When seeing these institutions named, one might think that examining a person’s fingerprint on social media is a sound method for assessing personality. However, as Brian Williams and Bill O’Reilly have recently demonstrated, just because something’s on the news doesn’t necessarily make it true.

In reality, this idea is just the latest chapter in an old work of fiction featuring things like tarot cards, astrology, crystal balls, graphology, and palm readings. It might be entertaining, but scientifically it’s a waste of time. Besides important privacy, professional, and legal concerns, there are a number of serious problems with using social media to measure a person’s personality. Here are just a few:

Some People Don’t Play the Game

A 2014 Business Insider article pegged the total number of Facebook users at 2.2 billion, or one-third of the global population. Although that number is impressive, it implies that two-thirds of the globe is not on Facebook. Some of these people are from impoverished countries without even basic human services; others are within our own borders, such as older Americans or those from socio-economic backgrounds not affording them easy access to the Internet. And yet others are well-educated adults who simply choose not to engage in social media. If social media is the future of personality measurement, should we just ignore all those people? Obviously not, but when one considers the fact that some organizations are using social media to screen applicants, this issue has huge implications.

Others Prefer to Watch

Like myself, many people on Facebook use the site not to post frequent messages, photos, videos, or other content about themselves, but to keep in touch with others by viewing what they post. I frequently visit Facebook, but use it to consume information rather than share it. Existing algorithms using Facebook Likes to assess personality are based only on information shared (and only a very specific subset of shared information at that), and thus have no means of describing such individuals.

You Are More than the Sum of Your Likes

Personality is far more complex than just the things you ‘Like’ on Facebook, and over-simplifying it only serves to dumb it down and make it less accurate. I’ll use myself as an example. Using the Cambridge website to describe me based on my Likes, it looks like I’m 29 years old (I’m 37) and likely masculine (62% probability; I’m 100% as far as I know), not married (27%; I should probably tell my wife), educated in Engineering (24%; Psychology was only 10%), and politically conservative (39%; definitely not).

Concerning my personality, my Facebook Likes paint me as Liberal and Artistic (wasn’t I also politically conservative?), shy and reserved, warm and cooperative, and calm and relaxed. If you don’t know me you might be tricked into thinking these are accurate. However, those who know me well would not be likely to describe me using these terms. For entertainment purposes, these inaccuracies are merely odd. However, considering the use of social media personality assessment for applied purposes, they have potentially disastrous consequences if used to screen job applicants.

Guessing < Basic Research

Let’s return to my Cambridge results describing me as a 29 year-old, politically conservative, single Engineer. These probabilities are not only inaccurate, but also easily corrected through publicly available data. If someone is sophisticated enough to check the About link on my Facebook page, they could see that I was born in 1977, earned a PhD in Industrial/Organizational Psychology, married my wife in 2004, and do not subscribe to either political party. This is a simple example, but underscores an important point – why use inaccurate probabilities to describe a person when better information is available? With demographics on Facebook, the About page is better than guessing. With personality measurement, well-constructed and validated personality instruments are preferred to Facebook Likes.

White-Collar and Blue-Collar Interests

Facebook Likes often reflect people’s personal interests and hobbies. It follows, then, that using Likes to measure personality will result in a profile based largely on assumed associations with those interests and hobbies. However, this logic is flawed. Previous research shows that blue-collar workers employed in occupations based in manual labor often pursue intellectual interests in their spare time to gain the intellectual stimulation that their jobs do not offer. White-collar workers with cognitively-based jobs, on the other hand, often engage in manual hobbies as a means of relaxing their minds in their spare time. These different personal interests are likely to be reflected in Facebook Likes, increasing the likelihood that personality assessment based on these data will be flawed and inaccurate.

Putting the Cart before the Horse

Although it’s interesting to think about how Facebook and other social media content reflects personality, those engaged in trying to use these data to predict personality have put the cart before the horse. In reality, clicking ‘Like’ on a topic on Facebook or otherwise engaging in social media is behavior, and like any behavior, it is our personality that predicts what we are likely to do when we engage in social media. Personality predicts behavior, not the other way around.

In this manner, those attempting to use Facebook to predict personality are much like other researchers who have examined the accuracy of job interviews in assessing candidate personality. Although responses to questions and behaviors demonstrated during the interview reflect a candidate’s personality to some degree, some personality characteristics (i.e., Extraversion) are easier to identify in interviews than others (i.e., Agreeableness). Put another way, interviews and social media presence may be interesting supplements to, but are no substitute for, directly measuring personality using well-constructed and properly validated assessment instrument.

OK, Wrap It Up…

Like the advent of the Internet itself, the rise of social media and the popularity of websites such as Facebook and Twitter have fundamentally changed how people communicate with each other and present themselves. Certainly, the content people share on social media reflects their personalities. But even at its best, trying to measure personality using algorithms that search content on the Internet is a poor substitute for using reliable and valid personality instruments. At its worst, particularly within occupational settings, such applications can be unprofessional, unethical, and unproductive.