Of course, you are a regular reader of Costa Connected, and have your privacy controls on Facebook pretty well locked down these days. You never accidentally share things publically that really ought to be kept for close friends, you don’t compromise or embarrass other people online. You aren’t leading a double life and you don’t have anything to hide, so why should you worry?
Well, sometimes we need to stop and think about the sheer volume of social signals we give off via Facebook, and ways that can be used to learn a staggering amount about you. Even before graph search made this information easily accessible to people you know, the aggregate data itself is so large that it can be used to predict things you may have specifically decided NOT to reveal on a public social network. When the data set of what you do share openly is large enough, it creates a window to the more private things.
Researchers in the US have been developing an algorithm that uses information based on the things people have ‘Liked’ on Facebook – media, groups, brands, businesses etc, the things we casually click on all the time – to try and draw conclusions about what they are really like. They tested their results using the profiles of over 58 thousand volunteers, who also completed details psychometric testing, to see how the ‘Likes’ predicted real psychological and behavioural traits. Their results, published in the Proceedings of the National Academy of Sciences, proved startlingly prophetic.
The algorithms emerged 88% accurate for determining male sexuality, 95% accurate in distinguishing black from white ethnicity, and also 85% accurate for indicating which of the two main US political parties respondents would vote for. And when it came to distinguishing men from women, 93% accuracy was obtained
Now, people use Facebook in different ways and some of the things you choose to like and share might very specifically indicate your political viewpoints, for example, because you want to publically identify with them and hope perhaps to influence others to feel the same. But other things are more personal, and out of the gay users surveyed less than 5% of them had ‘liked’ causes such as the campaign for gay marriage. The 88% correlation was drawn from analysis of likes of music, TV shows, news media and so on.
The implications of this are pretty chilling. You may not choose to declare your sexuality, or your religion, or your political preferences on Facebook – but if the right analysis is applied to your overall pattern of likes, these facts may be revealed. The volunteers in the test had an average number of 170 liked items, which isn’t even that high.
Of course for marketers this information is gold, because it means they can put offers and ads in the way of people who are most likely to respond. But it’s easy to forget there are real people with real lives behind the statistics, and the study references a US shopping giant which used customer shopping records to predict pregnancies of its customers, and send them carefully timed targeted offers. Now for most of the recipients the vouchers for prenatal vitamins and maternity clothes were probably very welcome, but if say their algorithm was 95% accurate – that’s 5% of very confused customers, and what about those who had just suffered a miscarriage, or if the mailshot revealed (or incorrectly suggested) pregnancy in an unmarried woman in a family/culture where this would be horribly unacceptable? This is where the boundaries of using generalised data can cross over into dangerous individual invasions of privacy.
Of course, storecard tracking data is commercially confidential to the particular store you have shared your ‘loyalty’ with – but Facebook ‘likes’ are by default public. And because of the simple ease of clicking on Facebook likes express a huge range of different, presumably positive, sentiment – I agree with that political statement, that cat looks nice, I also remember that ironically referenced product from the 80s, I find that sarky e-card mildly amusing, I would like people to think I am cool enough to identify with that brand which seems to be trending and also appears to be liked by several of my friends who I think are exemplars of cool themselves… So, we click away without a thought, for what that data could potentially reveal about us.
And what about those cool friends, whose association with brands is being used to make YOU more likely to click on that ad? Well, they might want to check their privacy settings. Personally I accept that Facebook now has hungry shareholders to feed, and they give me a great deal of stuff I like to free, so they have to serve me ads, that’s fine. What I don’t want is anyone seeing an ad accompanied by the line ‘Maya Middlemiss likes THIS’, not until Facebook is ready to pay me a royalty for the privilege – so to check your own settings for likes, visit this page https://www.facebook.com/settings?tab=ads§ion=social and opt to pair your ‘social actions’ (that’s liking stuff) with adverts for the eyes of – no-one.
Making your likes private in this way also hides your likes from pesky researchers too.
Published in Costa Blanca News, 5th April 2013