Is Facebook a Predictor of Your Health?


I’ve written before about the power of analyzing what you write in order to predict things about your health.  As more and more ‘big data’ companies and projects get publicity, it is fascinating to see how rapidly the field is growing.  We are at the point where each one of us experiences the web differently when we open our browser,  as our clicks and other data available on our habits and preferences are constantly analyzed. It’s all designed to get us to click or tap the ‘buy’ button.  I’d like to see some efforts made to use the same approach to motivate us to click/tap the ‘get healthier’ button.

With that in mind, I was fascinated to read a recent article in the Wall Street Journal (the news appeared in other outlets as well) citing a paper published in the Proceedings of the National Academy of the Sciences that studied how using the ‘like’ function on Facebook can reveal details about your personality.  Researchers first collected a lot of information from participants using standard personality inventories and psychological tests.  They then looked at patterns of clicking ‘like’ on Facebook to see if they could predict any personality traits or other identifiers.

The results were startling.  The researchers found, for example, that ‘likes’ for Austin, Texas, “Big Momma” movies, and the statement “Relationships Should Be Between Two People Not the Whole Universe” predicted drug use.  But “likes” for swimming, chocolate-chip cookie-dough ice cream and “Sliding On Floors with Your Socks On” were part of a pattern predicting that a person didn’t use drugs.  It gets better.  Patterns of using ‘like’ accurately distinguished between democrats and republicans 85% of the time, between black and white people in 95% of cases and between homosexual and heterosexual individuals 88% of the time.

You might be thinking that this is both technology and science run amuck.  “Why can’t you just ask me?” you might say.  It is worth pointing out the preponderance of evidence that, if asked about your health, you will predictably exaggerate those facts that make you look healthier and minimize those facts that make you look less healthy (this is predictable for some other types of questions too).  This is called “social desirability bias” and the phenomenon explains so much of what goes on both in the doctor’s office and when we talk casually to our friends and family about our health.

At the Center for Connected Health, we’re conducting a study now where we ask participants to fill out a short questionnaire on their readiness to move their behavior to a healthier state, followed by a period where we track their actual success (as measured by activity level).  Almost to a one, we’re finding participants overestimate where they are at the beginning of the study and need to be downgraded after one month of tracking.

There are many other examples of this phenomenon in the literature.  I find it liberating to think that there is a future (not too distant) where we’ll be able to calculate a health profile indirectly from records of your online and mobile behaviors (not just likes, but texts, emails, GPS data, etc.).  The ‘likes’ study is a great example of how that might work.

I can just hear the voices of the privacy advocates, the volume of their objections growing as they read this.  Rest assured, privacy will be preserved.  This health profile might be shared with you only, before it leaks to anyone else, even your doctor or your loved ones.  And if you object because it feels threatening to have someone tell you a version of the truth that might be too hard to hear, rest assured again.  A message that falls on deaf ears is of no real value in improving your health.  The message has to be customized to your state of readiness.  That is something we’re working on too.

At some point, there may be implications for your healthcare premium costs in this sort of model.  Exercising your libertarian right to privacy may lead you to have to choose higher insurance premium costs.  That is if we can show conclusively that this objectively-derived healthcare profile is accurate and predictive of your costs to the healthcare system.  There is a growing tension between ‘big data’ analytics and privacy.  When it’s designed to induce you to hit the ‘buy’ button, I think the argument to guard privacy is an easy one.  When we get it so that the design will be to induce you to improve your health, both for your own sake and for the sake of society, I’m not so sure.

What do you think?