After Cambridge Analytica, does Mark Zuckerberg’s Facebook know too much? - Times of London

march 21 2018, 12:01am, the times After Cambridge Analytica, does Mark Zuckerberg’s Facebook know too much? tom whipple, science editor Intelligent people like science, thunderstorms and curly fries. Less intelligent people like Harley Davidsons, the country music singer Lady Antebellum, and groups entitled “I Love Being a Mom”. The correlations between personality types and Facebook likes range from the blindingly obvious to the utterly cryptic. Gay men enjoy Wicked the Musical, but why do straight men like groups such as “Being Confused After Waking Up From Naps?” As far as data scientists, and people targeting election adverts, are concerned, it doesn’t matter. All that is important is what the numbers show. Using only your “likes” in 2013, two Cambridge scientists showed that they could predict people’s gender with 93 per cent accuracy, sexuality with 80 per cent accuracy and whether their parents were together at 21 with 60 per cent accuracy. In 2015, they showed that with 70 likes, they became better at predicting personality than a person’s friends while 300 likes gave scientists more knowledge about a subject than their own spouse had. The reason all this matters is that those scientists, Michal Kosinski and David Stillwell, are colleagues of Aleksandr Kogan, who this week was accused of passing data to Cambridge Analytica. On the face of it, this is troubling. Most of us are used to having targeted adverts follow us around the internet. But even so, there is something inherently creepy about the way in which apparently unconnected social media activity — following the Wu-Tang Clan, being interested in storms — can tell people we don’t know intimate facts about who we are. So, should we be worried? Is democracy dead? Not just yet, perhaps. The reason why, is that just because you can make a window into men’s souls doesn’t mean you know what to do with it. Cambridge Analytica may claim that it can use data to swing elections, but for a political campaign properly to use this information would require an advert tailored to each person. Alexander Nix, head of Cambridge Analytica, claimed that on one day in the Trump campaign they tested 175,000 different Facebook adverts. The Trump campaign dismissed that as a lie — but even if it wasn’t, how would we know they worked? The truth is, we wouldn’t. None of this has been tested. One form of personalised advertising has been investigated, however. Back in the pre-Facebook world, there was a way of pitching to each voter, using a semi-intelligent machine that appeared on your doorstep, responded to questions and changed its message accordingly. That machine was known as “a politician”. Recently, scientists investigated the efficacy of doorstepping. How many votes does it swing? As best they could estimate, none. Categorising minds may be hard; changing them is far harder.