Among the many threads, whispers, and rumors surrounding the election of US President Donald Trump, one in particular has caught the attention of the media and the public in recent weeks. In March, an exposé unveiled potentially illegal violations of online data privacy policies by Cambridge Analytica, a political analytics firm involved with several recent elections including the Brexit referendum and the 2016 US election. The firm used Facebook data to target potential voters. Today, questions are being asked about the security of our data, the proper role of social media in elections, and even the sanctity of the democratic process.
The origins of Cambridge Analytica can be traced back to Aleksandr Kogan, a researcher at the University of Cambridge, who used a personality quiz named ‘thisismydigitallife’ to harvest the data of 270,000 Facebook users and, crucially, limited data of all their friends, without their express permission. Through this, Kogan was able to access the information of over 87 million Facebook users, mostly in the United States and the United Kingdom.
Kogan, violating Facebook policy, transferred the data he had collected to SCL, the parent company of Cambridge Analytica. Kogan’s app remained on Facebook until 2015, when the social media platform requested Kogan and SCL to delete all of the data they had collected. Facebook, having created policies in 2014 that prevented that collection of a user’s friend’s data without permission and believing that the data had been deleted, chose not to alert those Facebook users whose data was accessed and did not take any further action.
The data that was gathered for Cambridge Analytica had immense value. Research done by Kogan and imitated by Cambridge Analytica focused on mapping the so-called ‘Big 5’ personality traits (openness to experience, conscientiousness, extraversion, agreeableness, and neuroticism) to social media activity through the analysis of a user’s Facebook likes, posts, and profile information. This work is part of an emerging field called psychometrics, or the quantitative analysis of psychological qualities. If successful, psychometric data can predict the psychology of any given user based on their Facebook profile. In a race to mobilise support, data such as this is of vital importance to political campaigns.
As a consulting and analytics firm, Cambridge Analytica used the information gathered to create a model that was able to predict qualities for over 200 million Americans based on publicly available social media data. Using this model and Facebook’s own ‘Custom Audience and Lookalike’ tools, Cambridge was able to target specific groups of users based on their social media activity. Firms like Cambridge Analytica were able to exploit Facebook’s data and advertising systems to ‘micro-target’ particular messages to those most vulnerable to them.
The ability to target voters with such specificity made Cambridge Analytica an attractive partner for many campaigns that were betting on strong voter data operations to win. Between 2015 and 2016, Cambridge Analytica partnered with US Senator Ted Cruz’s campaign for the Republican nomination for president, in which Cambridge’s efforts were described as the ‘key to his victory’ in crucial early primaries. Cambridge Analytica was also employed by the Leave campaign during the Brexit Referendum, Donald Trump’s presidential campaign, and other Republican campaigns across the US. For the Trump campaign, Cambridge’s model, despite purportedly not using Facebook data directly, was able to ‘identify which specific micro-targeted ads were trending in small geographic regions among specific personalities of Facebook users, and then direct the on-the-ground campaign to reinforce those messages in the form of a Trump stump speech.’ Seeing as how Trump spent considerably more time in key states down the stretch, it is possible to see how these techniques could have an important effect.
In the aftermath of the reports on Cambridge Analytica’s role in recent elections, there are two common responses that have circulated in the news. One focuses on the issue of privacy, calling for stricter regulations on how social media companies protect data. This is understandable given the wealth of information available about each person online and the way in which Cambridge Analytica was able to circumvent policies designed to protect us. The second focuses on the sanctity of our elections. Many see the advent of psychometrics as deleterious to the health of democracy; social media will now become a trojan horse for radical campaigns and foreign actors alike to seduce users with perfectly crafted messages. However, do these concerns matchup with the reality we are facing? Are failures of privacy regulations at fault for enabling Cambridge Analytica’s actions, and is the development of psychometrics a death sentence for social media and democracy? I argue no on both counts; the concerns over privacy and democratic health obscure from view the true enabling force: Facebook’s monetisation of user data. If people are truly concerned about psychometrics subverting democracy, they should look not at the acquisition of the data but at Facebook’s business model.
In response to the first concern, all of Cambridge Analytica’s actions were legal, and most did not violate Facebook policies. The only policy they did violate was the transfer of Facebook data from Aleksandr Kogan’s app to Cambridge for commercial purposes. The collection of the data in general, however, was allowed. While it is true that a new policy could have prevented Kogan gaining access to data of users’ friends, no amount of stricter privacy regulations or laws could have prevented Cambridge Analytica from gaining information on users since users themselves gave that permission. While an important question is how much information shouldbe hosted by social media platforms, privacy was not illegally thwarted by Kogan or Cambridge Analytica.
Furthermore, Cambridge Analytica did not need any user’s specific information to accurately target them. Through the collection of the data of 87 million users, Cambridge was able to construct a robust model generalisable to Facebook users as a whole. While Cambridge was asked to delete the data it had acquired, the models that were built off the data remained in their possession. These models have now been tested and refined through several campaigns throughout the United States and the UK. Since these models have been verified, any organisation with a large enough data file on potential voters, like the Republican National Committee, can successfully micro-target Facebook users, and no new privacy regulations can prevent that. To continue to operate, Cambridge Analytica would only need publicly available data to feed into their model, cutting against the image of an insidious hacker stealing the data of millions.
On the account of psychometrics heralding the corruption of democracy: psychometrics is only the latest in a line of similar tactics used widely by almost all political campaigns. In 2008, the Former President Barack Obama’s campaign was widely praised for innovative analytic techniques in which they assigned scores to every voter in the country based on how likely they would be to vote for Obama. This was so accurate that a campaign consultant claimed that ‘[w]e knew who…people were going to vote for before they decided.’ The Obama campaign then used this data to micro-target voters. In the digital age, every national campaign will have dedicated teams of analysts searching to maximise micro-targeting techniques.
The misplacing of concerns over information being stolen and democracy being subverted is also shown by the widespread use of psychometric tactics in the private sector. Before Aleksandr Kogan created his app on Facebook, his colleagues at Cambridge had already created an accurate algorithm that predicted personality traits, political tendencies, and even sexuality from Facebook likes. During the Iowa presidential primary elections in 2016, Dstillery, an analytics firm, used mobile phone location data to track over 16,000 voters. From this, they were able to compute similar connections to what would be found in Cambridge Analytica’s model. For example, they found that ‘people who loved to grill or work on their lawns overwhelmingly voted for Trump in Iowa.’ This company, like Cambridge Analytica, intends to sell their data and models to large firms to improve the advertising targeting methods popularly used on social media, and especially Facebook. Though perhaps not widely known, similar techniques to those used by Cambridge have been used for relatively benign purposes by private companies and political campaigns without the same uproar over privacy and democracy.
In the current narrative, all agency is bestowed to the analytics firms themselves as privacy violators and democratic hackers, with Facebook merely the means to an end. In all of this, however, Facebook is not just a passive tool; it is fundamentally organised to promote the active targeting of users. Concerns over privacy and democratic health, not reflecting the reality of the problem, conceal the real issue at hand: Facebook’s business model incentivising psychometric targeting.
Though Facebook markets itself as a social facilitator, it makes its money first and foremost from the sale of targeted ads. Users willingly tell Facebook what their favorite movies, books, and stores are, and in return Facebook serves advertisements designed to be the most relevant to their lives. Facebook has become one of the biggest and highest valued tech companies in the entire world by attracting companies to their advertising system, marketed as the most accurate ad platform on the internet. Facebook claims its targeting method is 89 percent accurate, compared to only 38 percent for Google’s AdSense. By this, Facebook’s business incentives allow campaigns to personally target users.
Further, the problem of campaigns using Facebook to mobilise support is exacerbated by the divisions which Facebook’s algorithm sows itself. Facebook prioritises posts on the news feed based on how much interaction they generate. The more interaction a post generates, the more engaged users are on the platform, and the more money Facebook can make from them. Thus, content that is divisive and controversial is favored. In this way, Facebook also derives benefit from the emotionally targeted content from groups like Cambridge through the generation of user engagement across the platform. Facebook both gives space to psychometric messages for a profit and amplifies their reach through content algorithms in order to maintain an engaged user base.
As long as Facebook continues to generate massive revenue through the sale of targeted advertisements, it seems unclear how to truly prevent organisations from micro-targeting consumers for their favorite soda or their least favorite politician. When questioned in front of the US Senate Judiciary and Commerce Committees, Facebook CEO Mark Zuckerberg said that while users always come before advertisers, he believes ‘that the ads model is the right one for [Facebook] because it aligns with our social mission of trying to connect everyone and bring the world closer together.’ Facebook also recently announced that consumers actually prefer to have relevant posts from companies on their news feeds and that they would not go ahead with separating promoted posts from the posts of friends and family. However, Facebook’s advertising model is the root of what users have taken issue. If the way Facebook monetises its service does not change, it is hard to expect interested groups to refrain from following the incentives Facebook promotes.
Given the influence of incentives driving the business models of social media platforms like Facebook, the superficial focus on privacy and the phantom concern for the health of democracy are only calling attention to two snowflakes amidst an avalanche. Stricter privacy regulations will not prevent organisations like Cambridge Analytica from continuing to produce divisive and emotionally targeted content. Only when the social media platforms themselves internalise their fundamental role in the proliferation of psychometric targeting in the public sphere can user concerns be totally redressed. Perhaps just as Upton Sinclair’s The Jungleunveiled the dirty underworld of the meatpacking industry, Cambridge Analytica can serve as an awakening to the men (and models) behind the curtain of social media.