Am 9. November gegen 8.30 Uhr erwacht Michal Kosinski in Zürich im Hotel Sunnehus. Der 34-jährige Forscher ist für einen Vortrag am Risikocenter der ETH angereist, zu einer Tagung über die Gefahren von Big Data und des sogenannten digitalen Umsturzes. Solche Vorträge hält Kosinski ständig, überall auf der Welt. Er ist ein führender Experte für […]
See also the reply from Carsten Drees: https://www.mobilegeeks.de/artikel/von-big-data-trump-und-bomben-ueberraschung-eure-daten-werden-genutzt/
(Automatic) translation with google translate:
I have only shown that there is the bomb
The psychologist Michal Kosinski has developed a method for analyzing human beings by means of their behavior on Facebook minutely. And helped Donald Trump win.
By Mikael Krogerus and Hannes Grassegger
The magazine N ° 48 – 3 December 2016
On November 9th, around 8.30 am, Michal Kosinski awakens in Zurich at the Hotel Sunnehus. The 34-year-old researcher has given a lecture at ETH’s risk center, a conference on the dangers of Big Data and the so-called digital revolution. Kosinski keeps these lectures constantly, all over the world. He is a leading expert on psychometrics, a data driven branch of psychology. When he turns to the TV, he sees that the bomb has burst: Contrary to projections of all leading statistician Donald J. Trump has been selected.
Kosinski long looked at the Trumps jubilee celebration and the election results of the individual federal states. He suspects that the result might have something to do with his research. Then he takes a deep breath and switches off the TV.
On the same day, a London-based London-based company, which is barely known, is sending a press release: “We are thrilled that our revolutionary approach to data-driven communications makes such a fundamental contribution to victory for Donald Trump,” an Alexander James Ashburner Nix is quoted. Nix is British, 41 years old and CEO of Cambridge Analytica. He always appears in the suit and with designer glasses, the slightly wavy blond hair combed backwards.
The thoughtful Kosinski, the locked Nix, the broad grinning Trump – one has made the digital revolution possible, one has succeeded him, one of them profited.
How dangerous is Big Data?
Anyone who has not lived on the moon for the last five years knows the term “big data”. Big Data also means that everything we do, whether on the net or outside, leaves digital traces. Any purchase with the card, any Google request, any move with the cell phone in the pocket, each like is saved. Especially everyone like. For a long time, it was not quite clear what these data should be good for – except that our blood pressure stenders are advertised in our Facebook feed because we have been grading “blood pressure lower”. It was also unclear whether Big Data is a great danger or a big gain for mankind. Since 9th November we know the answer. Because behind Trumps’s online campaign and also behind the Brexit campaign is one and the same big data company: Cambridge Analytica with its CEO Alexander Nix. Anyone who wants to understand the outcome of the election – and what Europe could be doing in the coming months – must start a remarkable incident at the British University in Cambridge in 2014. To Kosinskis Department of Psychometrics.
Psychometry, sometimes called psychography, is the scientific attempt to measure a person’s personality. In modern psychology the so-called Ocean method has become the standard. Two psychologists had succeeded in the 1980s to prove that every trait of a human being can be measured based on five personality dimensions, the Big Five: openness (how open you are to new things?), Conscientiousness (As you are a perfectionist?), Extraversion (How Sociable are you?), Tolerability (How ruthless and cooperative are you?) And neuroticism (Are you easily vulnerable?). Based on these dimensions can be relatively accurately say with what kind of people we’re dealing with, so that needs and fears he has, and also how it will behave tend. The problem for a long time was the data collection, because for the determination one had to fill a complicated, very personal questionnaire. Then came the Internet. And Facebook. And Kosinski.
Michal Kosinski, a student from Warsaw, began a new life when he was accepted at the venerable Cambridge University in England in 2008, at the Center for Psychometrics, at the Cavendish Laboratory, the first psychometrics lab at all. With a study colleague, Kosinski presented a small app to the then still manageable Facebook: On MyPersonality, the application was called, one could fill out a handful of psychological questions from the Ocean questionnaire (“Do you easily get stressed out of stress?” Do you tend to criticize others? “). As an evaluation one got his “personality profile” – own ocean values -, and the researchers got the valuable personal data. Instead, as expected, a few dozen study friends had quickly betrayed hundreds, thousands, and millions of their innermost convictions. Suddenly, the two doctoral students had the largest psychological record ever recorded.
The procedure that Kosinski develops with his colleagues over the next few years is actually quite simple. First, test persons are asked to submit a questionnaire. This is the online quiz. From their answers, the psychologists calculate the personal ocean values of the interviewees. Kosinskis team then resembles all possible other online scores of the test persons: what they have blogged, geshared or posted on Facebook, which gender, age, and where they have indicated. So the researchers get connections. From simple online actions amazing conclusions can be drawn. For example, men who liken the cosmetics brand MAC, with high probability gay. One of the best indicators of heterosexuality is the Liken by Wu-Tang Clan, a New York hip-hop group. Lady Gaga followers are very likely to be extroverted. Whoever is inspired by philosophy is more introverted.
Kosinski and his team refine the models unceasingly. In 2012, Kosinski proves that it is possible to predict from an average of 68 Facebook-Likes a skin color (95% accuracy) whether it is homosexual (88% probability) whether it is a Democrat or Republican (85%). But it goes even further: Intelligence, religious affiliation, alcohol, cigarette and drug consumption can be calculated. Even if the parents of a person have remained together until their 21st year of age, the data are shown. How good a model is is shown by how well it can predict how a test person will answer certain questions. Kosinski continues as if in an intoxication: Soon his model can estimate a person better than an average work colleague by means of ten Facebook-Likes. 70 Likes are enough to surpass the friendliness of a friend, 150 of the parents, with 300 Likes, the machine can predict the behavior of a person more clearly than their partner. And with even more Likes can even surpass what people believe of themselves to know. On the day when Kosinski publishes these findings, he receives two calls. A complaint and a job offer. Both of Facebook.
Only visible to friends
Facebook has now introduced the distinction between public and private. In the “private” mode, only your friends can see what you are talking about. But this is not an obstacle for data collectors: while Kosinski always requests the consent of the Facebook users, many online quiz today demand access to private data as a prerequisite for personality tests. (If you do not have much concern about your own data and want to let yourself be judged by their Likes on Facebook, this can be on Kosinskis page make applymagicsauce.com and compare then his results with those of a “classic” Ocean questionnaire: […]).
But it’s not just about the Likes on Facebook: Kosinski and his team can now assign people to the ocean criteria by their own portrait photo. Or by the number of our social media contacts (a good indicator of extraversion). But we also reveal something about us when we are offline. The movement sensor shows, for example, how quickly we move the phone or how far we are traveling (correlated with emotional instability). The smartphone, Kosinski notes, is a huge psychological questionnaire, which we constantly conscientiously and unconsciously fill out. Above all, and this is important to understand, it also works the other way around: you can not only create psychological profiles from data, you can also look for specific profiles – for example: all worried family fathers, all angry introverts. Or: all undecided democrats. What Kosinski has invented is a human search machine.
Kosinski is increasingly aware of the potential – but also the danger of his work.
The web always seemed to him like a gift from heaven. He really wants to return, share, share. Data can be copied, but all have some of it. It is the spirit of a whole generation, the beginning of a new age without the limits of the physical world. But what happens, Kosinski wonders, if someone abuses his human search machine to manipulate humans? He begins to warn all his scientific work. His methods could “threaten the welfare, freedom, or even the lives of men.” But nobody seems to understand what he means.
At this time, in early 2014, a young assistant professor named Aleksandr Kogan joined Kosinski. He had a question from a company interested in Kosinskis method. The Facebook profiles of ten million US users should be measured psychometrically. For what purpose, he could not say that there were strict secrecy requirements. Kosinski wants to say first, it is about a lot of money for his institute, but then hesitates. Finally, Kogan emerges with the name of the company: SCL – Strategic Communications Laboratories. Kosinski goes to the company: “We are a globally operating choice management agency,” he reads on the company website. SCL offer marketing based on a psychological model. Focus: Choice of influence. Election control? Disturbed to Kosinski click through the pages. What kind of company is it? And what do they have in the US?
What Kosinski does not know at this time: Behind SCL is a complex corporate design with taxpayers – as the Panama Papers and Wikileaks revelations show. Some have contributed to the turmoil in developing countries, while others have developed methods for the psychological manipulation of the population in Afghanistan. And now, SCL is also the parent company of Cambridge Analytica, the ominous big-data booth that organized the online election campaign for Trump and Brexit.
Kosinski does not know anything about it, but he does not know what is wrong. “The matter began to stink,” he recalls. In his research, he discovers that Aleksandr Kogan has secretly registered a company that does business with SCL. From a document, which is available to the magazine, it is clear that SCL was acquainted with Kosinskis method by Kogan. Suddenly Kosinski dawned that Kogan might have copied or rebuilt his Ocean model to sell it to the electoral control company. Immediately he breaks the contact to him and informs the institute leader. Within the university, a complicated conflict is developing. The institute is concerned about its reputation. Aleksandr Kogan first moves to Singapore, gets married and calls himself Dr. Specter. Michal Kosinski moves to Stanford University in the USA.
For a year, it is quite quiet, then, in November 2015, the more radical of the two Brexit campaigns, “leave.eu”, sponsored by Nigel Farage, announced that she had commissioned a big data company to support her election campaign online: Cambridge Analytica. Core competency of the company: novel political marketing, so-called microtargeting – based on the psychological ocean model.
Kosinski gets mails, what he has to do with it – with the keywords Cambridge, Ocean and Analytics many think of him first. For the first time he hears the company. He looks horrified at the site. His nightmare has come true: his methodology is used on a grand scale for political purposes.
After the Brexit in July, insults bellow at him: Just look what you have done, write friends and acquaintances. Everywhere Kosinski has to explain that he has nothing to do with this company.
First Brexit, then Trump
Ten months later. It is September 19, 2016, the US election is approaching. Guitar reefs meet the dark blue room of the New York Grand Hyatt Hotel, Creedence Clearwater Revival: «Bad Moon Rising». The Concordia Summit is a kind of global economy forum in Klein. Decision makers from all over the world are invited, among them there is also Federal Councilor Schneider-Ammann. “Please welcome Alexander Nix, Chief Executive Officer of Cambridge Analytica,” announces a gentle female voice from the off. A slim man in the dark suit enters the stage center. Silence prevails. Many here know: This is Trumps new digital man. “Soon you’ll call me Mr. Brexit,” Trump had tweeted somewhat a few weeks before. Although policy observers had referred to the similarity in content between Trumps Agenda and that of the right Brexit camp. However, few noticed the connection with Trumps’s recent commitment to a widely unknown marketing company: Cambridge Analytica.
Trumps digital campaign had more or less consisted of one person: Brad Parscale, a marketing entrepreneur and failed start-up founder, who had built a rudimentary website for 1500 dollars. The 70-year-old Trump is not a digital type, on his work desk is not even a computer. There is no such thing as an email from Trump, his personal assistant has once betrayed. She herself had persuaded him to a smartphone – from which he has been tweeting uncontrolled since then.
Hillary Clinton, however, relied on the legacy of the first social media president, Barack Obama. She had the address lists of the Democratic Party, gathered millions over the web, got support from Google and Dreamworks. When it became known in June 2016 that Trump Cambridge had hired Analytica, one’s nose was swept up in Washington. Foreign dudes in mass suits that do not understand country and people? Seriously?
“It is my privilege to speak before you, honored listeners, about the power of Big Data and the psychography in the election campaign.” Behind Alexander Nix, the logo of Cambridge Analytica appears – a brain composed of a few network nodes like one Map. “A few months ago, Cruz was still one of the less popular candidates,” says the blonde man with this British tongue, which makes Americans feel like many Swiss Germans, “only 40 percent of the voters knew his name.” Flash rise of conservative Senator Cruz. It was one of the strangest moments of the election campaign. Trumps, the last great opponent of the party, had come out of nowhere. “How did he do that?” Nix continues. By the end of 2014, Cambridge Analytica had entered the US campaign, initially as a consultant to the Republican Ted Cruz, financed by the secret US software billionaire Robert Mercer. So far, according to Nix, electoral campaigns had been conducted according to demographic concepts, “a ridiculous idea when you think about it: all women get the same news, just because they have the same sex – or all African Americans because of their race?” The Campaign team of Hillary Clinton, does not need to mention anything here, it divides the population into supposedly homogeneous groups – just as all the opinion research institutes did, which Clinton saw until recently as a winner.
Instead, Nix clicks on the next slide: five different faces, each face corresponds to a personality profile. It is the Ocean model. “We at Cambridge Analytica,” says Nix, “have developed a model that can calculate the personality of every adult in the US.” Now it is absolutely silent in the hall. The success of marketing by Cambridge Analytica is based on the combination of three elements: psychological behavioral analysis according to the Ocean model, big-data evaluation and ad-targeting. Ad targeting, that is personalized advertising, that is advertising, which adapts as closely as possible to the character of a single consumer.
Nix frankly explains how his company does it (the lecture is available on Youtube free). Cambridge Analytica purchases personal data from all sorts of sources: land registry entries, bonus cards, dial-up directories, club memberships, journal subscriptions, medical data. Nix shows the logos of globally operating data traders like Acxiom and Experian – in the USA almost all personal data are available for purchase. If you want to know where, for example, Jewish women live, you can simply buy this information. Including telephone numbers. Now Cambridge Analytica crosses this number packets with voter lists of the Republican Party and online data as Facebook Likes – then one calculates the Ocean Personality Profile: From digital footprints suddenly real people with fears, needs, interests – and with a residential address.
The procedure is identical to the models developed by Michal Kosinski. Cambridge Analytica also uses IQ-Quiz and other small Ocean-Test apps to get to the meaningful Facebook-Likes of users. And Cambridge Analytica does precisely what Kosinski had warned: “We have psychograms from all adult US citizens – 220 million people,” Nix opens the screenshot, “our control centers look like. Let me show what we do with it. “A digital cockpit appears. Left-hand diagrams, on the right a map of Iowa, where Cruz had collected a surprisingly large number of votes in the pre-election campaign. Thereupon hundreds of thousands of small dots, red and blue. Nix borders the criteria: Republicans – the blue dots disappear; “Not yet convinced” -points disappear again; “Male” and so on. At the end, a single name appears, including age, address, interests, political inclination. How does Cambridge Analytica now handle such a person with political messages?
In another presentation, Nix shows two versions of how to appeal to psychologically ill-informed voters: “For a fearful person with high neuroticism values, we sell the weapon as an insurance. “The right side shows a man and a child in the sunset, both with rifles in a field, evidently in the duck hunts:” This is for high-class conservatives Extraversion »
How to keep Clinton voters away from the urn
Trumps of conspicuous contradictions, his often criticized attitude and the resulting immense amount of different messages suddenly turn out to be his great advantage: each voter his message. “Trump acts like a perfectly opportunistic algorithm, which only depends on audience reactions,” notes the mathematician Cathy O’Neil in August. On the day of the third presidential debate between Trump and Clinton, Trumps Team sends 175,000 different variations of its arguments, mainly via Facebook. The messages differ only in microscopic details, in order to suit the recipients psychologically optimally: different titles, colors, subtitles, with photo or with video. The fineness of the adaptation goes down to small groups, explains Nix in conversation with «Das Magazin». “We can reach villages or blocks of houses in a targeted manner. Even individuals. “In Miami’s Little Haiti district, Cambridge Analytica provided residents with news of the Clinton Foundation’s failure after the earthquake in Haiti – to stop them from choosing Clinton. This is one of the goals: potential Clinton voters – this include dubious leftists, African Americans, young women – from the urn, “oppress their choice,” as a Trump associate tells. In so-called dark posts that are purchased Facebook ads in the timeline that can see only users with a matching profile, videos, for example, African Americans leaked in which Hillary Clinton black men called predators.
“My children,” Nix concludes his lecture at the Concordia Summit, “will no longer be able to explain something like an advertising poster with the same news for all, indeed the whole concept of a mass media. I thank you for your attention and can tell you that we are now working for one of the two remaining candidates. “Then he leaves the stage.
How precisely the American population is massaged at this moment by Trumps digital troops is not recognizable – because they seldom attack broadly in the mainstream TV, but mostly personalized on social media or in digital television. And while the Clinton team is safe on the basis of demographic projections, in the headquarters of the Trump digital campaign, a “second headquarters” is emerging in San Antonio, as Bloomberg journalist Sasha Issenberg notes with surprise after a visit. The Cambridge Analytica team, allegedly a dozen people, had received about $ 100,000 from Trump in July, $ 250,000 in August, and $ 5 million in September. All in all, says Nix, you have taken about 15 million dollars.
And the company’s measures are radical: starting in July 2016 Trump-Wahlhelfer will provide an app with which they can recognize the political attitude and the type of personality the inhabitants of a house have. When trumps are ringing people at the door, only those who call the app as receptive to their messages. The electoral assistants have adapted the guide for the personality of the resident. The response, in turn, gives the appointees to the app – and the new data flows back into the control room of Cambridge Analytica.
The company divides the US population into 32 personality types, concentrating only on 17 states. And as Kosinski had found that men, the MAC Cosmetic liken, are very likely gay, Cambridge Analytica found out that a preference for US-made cars is the best sign of possible trump voters. Among other things, Trump now shows which messages are moving and where exactly best. The decision to focus on Michigan and Wisconsin in recent weeks is based on data analysis. The candidate becomes a model of the implementation of a model.
What is Cambridge Analytica doing in Europe?
But how great was the influence of psychometric methods on the outcome of the election? Cambridge Analytica does not want to provide evidence on the effectiveness of the campaign. And it is quite possible that the question can not be answered. And yet, there is evidence: the fact that Ted Cruz, thanks to the help of Cambridge Analytica, rose from nowhere to the strongest competitor Trumps in the primaries. There is the increase of rural electorate. There is the decline of the votes cast by African Americans. The fact that Trump spent so little money could also explain the effectiveness of personality-based advertising. And also that he put three quarters of his marketing budget in the digital area. Facebook proved to be the ultimate weapon and the best choice, as a Trump staff tweaked. In Germany, for example, the AfD, which has more Facebook friends than the CDU and the SPD.
It is by no means so often said that statisticians have lost this election because they are so close to their polls. The opposite is true: the statisticians have won the election. But only those with the new method. It is a strolling joke of the story that Trump often criticized the science, but probably won the election thanks to her.
Another great winner is Cambridge Analytica. Its CEO, Steve Bannon, publisher of the ultra-right online newspaper “Breitbart News”, has just been appointed to Donald Trumps as chief strategist. Marion Le Pen, from the French Front National, has already tweeted the fact that she has accepted his invitation to work together, and on an internal corporate video is talking about the recording of a «Italy» meeting. Alexander Nix confirms that he is on customer acquisition, worldwide. There were inquiries from Switzerland and Germany.
All this Kosinski has observed from his office in Stanford. After the US election the university is headed. Kosinski responds to developments with the sharpest weapon available to a researcher: with a scientific analysis. Together with his research fellow, Sandra Matz, he has conducted a series of tests, which will be published soon. First results available to the magazine are disturbing: psychological targeting, as Cambridge Analytica used, increases the click rates of Facebook ads by over 60 per cent. The so-called conversion rate, that is, how strongly people – after they have seen the personally tailored advertising – also act afterwards, ie make a purchase or even choose, increased by an incredible 1400 percent.
The world has turned. The British leave the EU, in America Donald Trump reigns. It all started with a man who wanted to warn us of the danger. At the now again these Mails arrive, which accuse him. “No,” Kosinski says quietly, shaking his head, “this is not my fault. I did not build the bomb. I have only shown that they exist. ”