The Cambridge Analytica scandal rocked the world a few months ago. But WTF did they do?
At the center of the whole controversy was Facebook & how Cambridge Analytica broke data privacy rules & got access to a ton of personal data. The main issue was that Cambridge Analytica got hold of personal data of around 87 million Facebook users. They collected this data via a Facebook quiz app called ’This is your digital life’, but also exploited a loophole that allowed it to collect data from the Facebook friends of the quiz takers as well.
But what did Cambridge Analytica really do with this data, and why was it deemed so powerful? What they (apparently) did might be in the grey regions of ethics but it was really innovative — involving the dark arts of machine learning. Essentially Cambridge Analytica is a big data company. They used this vast personal data to create psychometric models that could predict character traits of any person by just using their Facebook likes!
To get an idea of just how accurately they could predict, read the excerpt below from an absolutely recommeded article on this issue, ‘The Data That Turned the World Upside Down':
In 2012, … proved that on the basis of an average of 68 Facebook “likes” by a user, it was possible to predict their skin color (with 95 percent accuracy), their sexual orientation (88 percent accuracy), and their affiliation to the Democratic or Republican party (85 percent). But it didn’t stop there. Intelligence, religious affiliation, as well as alcohol, cigarette and drug use, could all be determined
Psychographic analysis has been around for a long time now. The models they used, called ‘Big Five’ or OCEAN, was actually developed in the 1980s to ‘measure’ human being’s characteristics based on five personality traits. But what changed now was the availability of big data. Their researchers now had enough personal data from millions of people to be able to perfect (train) the OCEAN model & improve its prediction accuracy.
But how did they achieve these amazing prediction models?
First, they had millions of people fill up online questionnaires (in the form of a Facebook quiz) that allowed them to calculate the OCEAN values for each respondent. Then they correlated these traits with their online behaviour on Facebook — likes, shares etc. They then established strong correlations between certain online actions such as ‘liking certain brands’ & character traits such as introvert or extrovert. Once you have these correlations in a model, & you know a few ‘likes’ of a random person, you can basically trace those ‘likes’ to character traits. Repeat this for millions of data points, and you have frighteningly accurate prediction models.
And why psychometric (or psychographic) segmentation? Because if you know the behaviours, opinions, values & feelings of your customers, you can personalise & micro-target your customers.
— — — — — —
Get these daily blogs in your inbox. Subscribe now!
From the desk of Aditya Khanduri
About me (Why I started Polygyan)
