Big data and psychographic profiling helped Donald Trump win the US presidential election

Aditya MadanapalleJan 31, 2017 11:45:38 IST

Contrary to expectations from polls, Donald Trump secured a surprise victory in the 2016 US Presidential Election. After the victory, there has been a scramble to explain the results.

Echo chambers on social media platforms where biased misinformation were circulated were speculated to have contributed to the surprise win. Barack Obama accused Facebook of allowing fake news stories to spread on the platform, which helped Trump win.

Facebook denied the allegations, but took strong steps to restrict the spread of misinformation on the social network. Donald Trump agreed that social networking platforms such as Facebook and Twitter did help him win the elections. The resulting picture is that hordes of fans spread pro Trump messages and silently voted for Trump in the election booth without publicly displaying their affiliations, even to friends and family.

An article on Motherboard suggests that big data, psychographic profiling and targeted content was effectively used by the Trump campaign.

Image: Reuters

So far, the traditional approach for advertising has been as a mass communication effort, where bright minds gather around and think of slogans that people can associate with and flock to. New tools and platforms available today allow for dissemination of messages that can resonate much more personally with each individual.

Demographic profiling assumes that a message will appeal to an intended cross section of the audience, based on race, gender or age. While demographic profiling can work to an extent, there is another, more effective tool, that allows campaign organisers to know what kind of messages work with which individuals, based on the psychologies of the people.

The base psychographic model for a profile is an assessment based on five personality traits, known as 'Ocean'. The openness, contentiousness, extroversion, agreeableness and neuroticism of individuals are measured. The model is known as the "big five" and has been around since the 1980s. Analysts can understand the needs and fears of the people based on these traits, as well as how they are likely to react to certain messages.

Image: Cambridge

However, collection of the data has been a problem so far, because building the profiles required a long, highly personal questionnaire. This is where the Internet and social media platforms come into the picture.

Michal Kosinski and David Stillwell were students at the Cambridge University Psychometrics Centre. They introduced an app on Facebook in 2008 that allowed users to take a quiz to understand and share their own psychographic profiles, and expected a few dozen responses from colleagues and friends.

Instead, they got over a million responses, and had the largest psychographic database linked to Facebook accounts ever. Kosinski correlated the results of the psychographic profiles to the likes of the respondents, and was able to make deductions based on these correlations. Cambridge has made available the API it developed for translating digital footprints into psychographic profiles available publicly for researchers.

Users who liked posts related to Lady Gaga were more likely to be extroverts. Hillary Clinton smiles while being introduced with Lady Gaga at a campaign rally in North Carolina November 8, 2016. Image: Reuters.

For example, those who liked Lady Gaga were more likely to be extroverts, men who liked the cosmetics brand MAC were more likely to be homosexual, those who liked philosophy related posts were more likely to be introverts, and liking the Wu-Tang clan was a strong indication that the individual was heterosexual. Each correlation or data point was too weak on its own to make an accurate prediction, but thousands of data points combined allowed for increasingly accurate predictions. Users can test out the accuracy of the profiling for themselves on Kosinski's site.

Soon Kosinksi was able to assess and predict an individual's profile based only on Facebook likes, and with a chilling degree of accuracy. From 70 likes, Kosinski knew more about the person than his friends. He could predict the skin colour, sexual orientation and political affiliations with greater than 85 percent accuracy. At 150 likes, he could better predict the person than the parents of the individual, at 300, he knew better than the partners. More likes, and Kosinski could know more about the person than the person knew about himself.

Kosinski was given a job offer by an assistant professor on behalf of a company known as Cambridge Analytica. Kosinski did not take up the offer, but suspects that Cambridge Analytica duplicated his approaches. Cambridge Analytica is a spin off of SCL Group, a company that claims to work at the forefront of behavioral change communication.

SCL Group is involved in commercial psychographic targeting for advertisers, social campaigns for NGOs, conflict resolutions through information dissemination, and offers behavioral modelling along with micro-targeting services for political campaigns.

They will soon be calling me MR. BREXIT!

— Donald J. Trump (@realDonaldTrump) August 18, 2016

Some commentators noticed the parallels in the Brexit movement and the alt-right organisations that were vocal supporters of Trump. Cambridge Analytica has not commented on its involvement with the Brexit campaign.

However, it does claim to have provided Donald Trump with the expertise and intelligence required to win the White House. Cambridge Analytica claims to have provided the right messages targeted to the right people at just the right time to turn around the voters.

Cambridge Analytica claims to have psychographic profiles of all the 220 million adults in the United States. The team included data scientists with PhDs, digital marketing strategists as well as content creators.

The messaging based on the insights provided by the data were thoroughly tested, and delivered through targeted posts in key battleground states. There are adjustments made on a daily basis, to make sure that the targeting and communication is most effective at any point of time.

Over 20 custom models were built for the Trump campaign. The models assessed voters based on policy issues, candidate preference, likelihood of voting early, and most significantly, identified the voters who were most likely to be persuaded.

The campaign was fine-tuned to target specific messages at particular psychographic profiles, messages that would not publicly appear to everyone. Individually targeted communication was chosen over mass communication. Here is Aleksandr Nix, CEO of Cambridge Analytica explaining the approach at the 2016 Concordia Summit.

Psychographic profiling was used for "dark posts" on Facebook, that only the targeted could view. A neighborhood or village could be targeted, down to a single individual. African american men for example, were shown videos where Hillary Clinton refers to black men as predators.

Cambridge Analytica claims to not use data from Facebook. Kosinski is troubled because of the impact of the activities of Cambridge Analytica can be associated with his work at Cambridge. Kosinski plans to fight back against the companies that abuse psychographic profiling and associated data with more data science, publishing results on the chilling effectiveness of personality targeting.