What is the science behind the Cambridge Analytica scandal?

Open this photo in gallery:

On one hand, this week’s Cambridge Analytica-Facebook fiasco seems like a new kind of story, one that could only have arisen like Frankenstein’s monster out of the data-driven hyper-connected reality our world has become.

But underlying the technology is a principle so old it is still commonly referred to by its Latin name: caveat emptor. Let the buyer beware.

This puts the responsibility squarely on the buyer to check out the goods before a purchase is made. It also reminds us to beware of advertisements that promise but may also mislead. The new twist that has emerged from the marriage of psychology with social media and big data is that we, the buyers, must now beware of ourselves. In the brave new world of data analytics, it is our own innate characteristics and behavioural profiles that can be turned against us by unscrupulous actors.

“It’s really the unique individualization that makes this different,” said Jacob Hirsh, a psychologist who specializes in organizational behaviour at the University of Toronto, Mississauga. “Behavioural profiles are all about the individual, and based on every online digital footprint that is available to disclose aspects of that individual.”

For years, researchers have understood that human behaviour tends to match up with personality traits. The so-called “big five” traits, openness, conscientiousness, extraversion, agreeableness and neuroticism (known collectively by the acronym OCEAN) form a continuum against which an individual can be scored and identified. This specificity creates an opportunity for targeted marketing.

In 2012, for example, Dr. Hirsh and colleagues published a study in which 324 survey respondents were shown advertisements for a fictitious smartphone in which the wording of the ads was chosen based on their personality scores. Those who scored high on openness received an ad that pitched the phone as a tool for accessing new applications and experiences. Those who scored high on neuroticism saw an ad that stressed security first. The results showed that the personality matching was effective at increasing the impact of the ads.

That study was referenced in later work by researchers at Cambridge University who used personality tests from tens of thousands of volunteers, administered through Facebook, to show that personality traits also correspond to a user’s Facebook “likes.” The team then developed a computer model that could reverse-engineer the process, using “likes” to predict personality. In 2015, the researchers showed that the model was more accurate than a human at judging personality and predicting political attitudes.

So far so academic. But Facebook’s vast user base offered an unprecedented opportunity to apply the model at an even larger scale.

“That’s where the power of big data lies,” said Murat Kristal, director of a business analytics program at York University’s Schulich School of Business in Toronto.

Enter Cambridge Analytica, the company co-founded by Steve Bannon, former adviser to U.S. President Donald Trump, where analysts repurposed the university-based research into a tool for political messaging. To do this, the company exploited a loophole that allowed access not only to the “likes” and other information from 270,000 people who participated in a survey, but the preferences of their Facebook friends as well. In the end, personality-relevant data were gathered from some 50 million individuals, equal to more than one-third of the number of people who voted in the 2016 U.S. presidential election. Each individual was a potential target for tailor-made messages relating to the vote.

Yet opinions vary as to how much of an impact the data made. The personality traits the company was trying to target are somewhat fluid and can shift from day to day around an individual’s average. Ultimately, “there are going to be many factors that influence the political choices people make,” said Tera Letzring, a personality researcher at the Idaho State University.

Nevertheless, in a close election where some swing states were won by thousands, not millions of votes, it’s possible to imagine a small effect making a big difference. That makes this week’s revelations a wake-up call, both for consumers and those aiming to reach them.

“The providers of analytics have to be responsible,” said Jan Kestle, president and CEO of Environics Analytics, a marketing and analytical services company, “We have to answer the question not only what can we do, but what should we do?”

Editor’s note: Tera Letzring is a personality researcher at the Idaho State University, not the University of Idaho.