The recent revelations involving Cambridge Analytica and Facebook portrayed a genius plan foiled by the perpetrators’ own, grave misunderstanding of human nature. Cambridge Analytica’s intention was to use the innocent cover of “academic research” to syphon vast amounts of data from both compliant and unknowing Facebook users. The cover of research was provided by Cambridge University’s Aleksandr Kogan, who offered Facebook users money to fill in a personality survey online. The survey responses and personal attributes, such as the “likes”, gender, sexual orientation, relationship status, and education of the 270,000 respondents and their friends amounted to a dataset involving roughly 50,000,000 Facebook users. The company then used the data to build a model that could predict people’s personalities relying on their Facebook profiles only. According to Cambridge Analytica’s own advertising, their model’s predictions about their targets’ personalities outperformed even close friends and family. Cambridge Analytica’s business model was to use their newly created tool to identify Facebook users with the right personality and to serve them micro-targeted advertising to further customers’ political agendas. Cambridge Analytica’s customers came from conservative circles and included the instigators of Brexit, the campaign of Donald Trump, and the advisers around Ted Cruz. If both Kogan and Cambridge Analytica had not gotten one key detail wrong, they may have built the machinery that could have accelerated further the political unraveling we are currently experiencing in the United States and European Union.
While Cambridge Analytica’s model excelled in predicting targets’ big five personality traits, openness, conscientiousness, extraversion, agreeableness, and neuroticism, or OCEAN for short, the company’s capabilities to target people’s political preferences seemed to lag behind the expectations. Feedback from Ted Cruz’ campaign provided evidence of that, and so did the revelation that Cambridge Analytica’s involvement in Trump’s campaign had been overstated by the company’s CEO. Interestingly, Cambridge Analytica also offered other, non-data-driven approaches for political coercion, such as extortion, honey-traps, and smear-campaigns, as documented by a secret video taken by UK’s Channel 4. Their willingness to add much less subtle, yet equally dirty tools to their offerings can be interpreted as a red flag for the efficacy of their data-driven approach. If Cambridge Analytica tried to create a political bomb by designing a sophisticated and quite creative big-data personality model, then why did it misfire? I would argue that creating the tool they envisioned takes an understanding of how personal views and political preferences evolve, which they did not have.
Early on, political scientists doubted that personality traits could be an accurate indicator for the political leanings of people. Are there more conservative or progressive extroverts? The correct answer to this question may be elusive, as there might not be one; The OCEAN-traits are likely useless in determining motivations, preferences, goals, incentives, or political leanings in people. Even a handful of additional traits, such as gender, sexual orientation, age, job, university major, and IQ, which Cambridge Analytica added to supplement the OCEAN-traits in their model, may have not changed the accuracy of their tool in predicting political leanings significantly. So what may be the common denominator among political leanings? Personal values.
Imagine personal values to resemble colored eyeglasses that every single one of us looks through to see the world. For almost all of us, it is impossible to see the glasses we are constantly staring through. Instead, we see differently-colored glasses on people we encounter in daily life, but we do not deduce from it that we are wearing colored glasses ourselves. Rather, we believe that there are some people with glasses of interesting colors, and some people who do not wear glasses, ourselves included. However, instead of choosing a pair of glasses randomly, at birth perhaps, there is a progression in the colors of glasses we put on throughout our lives. Each successive pair of glasses does not fully replace the previous one, but it brings a wider field of our vision into focus. Therefore, sets of values we acquire later in life are not superior, but simply more adapted to the increasingly complex situations that we encounter as we age. As a baby, we have a very specific color of shades on our noses, which only lets us see, and value, very basic survival needs: discomfort and hunger. Thus, we value comfort and dislike pain, and we value food and dislike hunger. This set of values is perfectly acceptable for a newborn, but as soon as babies start to see faces and recognize them, a more complex set of values is necessary to distinguish friend and foe. Who is part of my family, and who is not? At that stage, children will dislike being held by strangers, as they now value the safety associated with familiar faces. It is important to point out, however, that these more sophisticated values transcend and include previous ones, as hunger and comfort is still important at that stage, and, as most of us will attest to, at much later ages as well. Once food and comfort are not secured anymore, a child’s fear of strangers may vanish and they may take food from anybody who will offer it to them. Through our entire lives, these changes in colored glasses, or values, will happen naturally, sometimes without us noticing, and sometimes through violent phases of struggle, inner conflict, depression, or pain. Another example for the evolution of values that any person may go through, if circumstances permit, is the transition of some of the world’s most successful businessman, such as Bill Gates and Warren Buffet. After a period of aggressive money-making, during which they achieved fame, money, acknowledgement, and status, a more humanitarian set of values emerged in both of them, which let their main focus be on charity and the people around them. This example also suggests that political values are tightly linked to, or even identical to a person’s “core values”. Someone who holds a humanitarian set of values will automatically lean towards more progressive, human-focused political parties, whereas people who are struggling to secure money and gain status may remain more focused on ethnocentric or personal gains, thus being more in line with conservative, or even “right-wing” parties. This general lack of differentiation between personalities and values made Cambridge Analytica’s tool so imperfect, despite their astonishing accuracy in predicting personality traits.
Approximately one year ago, a friend and I started to venture into a project aiming at building value-assessment tools. Early on, we started considering using data from Facebook users in conjunction with our value questionnaire to connect values and Facebook profiles. In light of the recent Cambridge Analytica scandal, this may cause some uneasiness in a lot of the readers, and I can fully empathize with them. The power of such a tool in marketing and politics would be immense. However, so would its power in the use we intended for it, which is mediation, human resources, and tackling the challenges of digitalisation in the 21st century. As with all technology, the motivations - or let’s call it by its name - the values, make all the difference whether a technology becomes a tool for humanity or a weapon against it. In light of the Cambridge Analytica scandal, we can only be glad that the people working on the technology did not have the capacity to reflect upon their own values, thus being unaware of the topic and choosing the wrong metric for a technology that they intended to weaponize early on. It is important that we, as the general population, are aware of this topic and the existence of different value-systems, so we can be alert when the next company with malicious intentions, or rather ethnocentric, self-centered, power-hungry values, gets “it” right.