A new study casts new light on how social networks can gather information about you — even if you don’t have an account.

In a way, social media is like smoking — but instead of being bad for your health, it’s bad for your privacy. There’s another striking similarity between the two: just like second-hand smoke is a thing, affecting those who might not even smoke, social media might also affect the privacy of those around you, even if they’re not users themselves.

The new study from researchers at the University of Vermont and the University of Adelaide gathered more than thirty million public posts on Twitter from 13,905 users.

The first concerning find is that it only takes 8 or 9 messages from a person’s contacts to be able to predict that person’s later tweets “as accurately as if they were looking directly at that person’s own Twitter feed”. In other words, social media information about yourself can also be derived indirectly.

“You alone don’t control your privacy on social media platforms,” says UVM professor Jim Bagrow, one of the authors of the study. “Your friends have a say too.”

“You think you’re giving up your information, but you’re giving up your friends’ information too!” adds University of Vermont mathematician James Bagrow who led the new research.

UVM professor Jim Bagrow led a new study, published in Nature Human Behavior, that suggests privacy on social media networks is largely controlled by your friends. Image credits: Joshua Brown.

The study also found that if a person leaves social media (or never joined it in the first place), 95% of this predictive accuracy also stands. Scientists found that they were generally successful at predicting a person’s identity and future activities even without any data from them.

This raises fundamental questions about how privacy can be protected. Intuitively, you would think that if you’re not on a social network, nothing can be known about you.

However, scientists have also shown that there is a fundamental limit to how much predictability can come with this type of data.

“Due to the social flow of information, we estimate that approximately 95% of the potential predictive accuracy attainable for an individual is available within the social ties of that individual only, without requiring the individual’s data,” researchers conclude.

The study has been published in Nature