Screenshot 2017-02-03 at 3.30.18 PM.png

Psychometric profiling mines big data from social media platforms to create advertising tailored to the personality traits of select people. A screenshot of Cambridge Analytica’s Data Dashboard tool, which provides demographic data based on the OCEAN personality model to political campaign workers (via cambridgeanalytica.com)

 

This sort of data analysis always freaks me out.

 

You’ve probably noticed that the ads which pop up on your browser and Facebook feed are highly relevant to you and often feature products you’ve purchased in the past. You may already know that this is because of your digital footprint-the trace you leave when you visit a web page or use your credit card to buy something. How does the internet know this about you? The answer lies in big data and the world of statistical programming. Statistical programming is a way to mine extremely large amounts of data for predictive modelling. Computer programs use complicated mathematics to analyze volumes of data too big for the human mind.

 

Predicting climate and weather patterns is one form of statistical modelling. An enormous amount of data on temperature, humidity, and wind, among other variables, are analyzed by computer programs which then generate predictions of future climate patterns. Another, based on marketing, has recently been developed using data from Facebook.

 

Begun in 2008 by then-doctoral student Michal Kosinski while at Cambridge University, the project aimed to measure anyone’s personality according to five traits psychologists term OCEAN-openness, conscientiousness, extraversion, and neuroticism. How much you enjoy new things, how much you care about taking care of someone else’s needs, how much you like to spend time with others, and what kind of anxious tendencies you have. These traits are remarkably accurate in how they can predict behavior. What Kosinski did was figure out a way to assess someone’s OCEAN profile based entirely on their Facebook activity. He started by sending out questionnaires to friends.

 

The results were then compared with their Facebook activity-what they liked, posted and shared. As Facebook grew, so did the pool of questionnaires and profiles. Very strong correlations between respondents’ questionnaires and their Facebook activity emerged. Gay men are more likely to ‘like’ the cosmetic line MAC. Straight men are more likely to ‘like’ Wu-Tang Clan. By 2012, Kosinski’s team was able to predict age, skin color, religious and political affiliation, and many other traits, from 68 likes on Facebook.

 

Well, so what? It turns out that you can do a lot with this information, as Kosinski’s team discovered when they were approached by private firm Cambridge Analytica with an offer to purchase usage rights of the research. Cambridge Analytica designed models for engaging with different OCEAN types and developed marketing to appeal to someone based on those traits.

 

Guess who hired Cambridge Analytica for targeted marketing? Both the Brexit and Trump campaigns. While Kosinski claims that it’s impossible to know how much his research affected election outcomes, one thing is certain: there’s going to be a lot more targeted marketing in the coming years.

 

Have a story tip? Message me at: cabe(at)element14(dot)com

http://twitter.com/Cabe_Atwell