The digital revolution initiated by Michal Kosinski at the Psychometric Centre at the University of Cambridge is being continued by a team of experts led by young development director Vesselin Popov. The Centre uses algorithms to analyze personal data that you leave on the internet – everything you have ‘liked’ on Facebook, posted on Instagram, ‘googled’, or purchased in e-commerce – and constructs your personality, creating precise psychodemographic profiles in a fraction of a second. Such analyses hold great value for marketing agencies and brands striving to maximize product personalization to achieve higher conversion rates. It is a public secret that these technologies are also used in political spheres, where politicians tailor their rhetoric to gain as many voters as possible. Kosinski’s algorithm, for example, was utilized by Strategic Communication Laboratories (SCL). When he suspected that his achievements would be used in electoral campaigns, he left that business, and SCL founded a new company specializing in American elections – Cambridge Analytica, which is currently at the center of a major scandal for using the private data of 50 million Facebook users.
>>>Snowden on the ‘most successful deception’: Facebook is a surveillance tool rebranded as a social network
We spoke with Popov about the role of psychometrics in Trump’s victory, Brexit, the potential use of predictive technologies in business, and the future of privacy rights.
- What are you currently working on at the Psychometric Centre and what is your role in it?
– We study human behavior and motivation using large amounts of data from social networks, online job boards, dating sites, the Internet of Things, and similar sources. Our research is focused on developing tools and methodologies for psychological and behavioral assessment, and my role is to identify areas where psychometrics could be applied, whether it is optimizing selection processes, personalizing digital services, or improving the quality of healthcare.
- You coordinate the prediction algorithm Apply Magic Sauce. How does it differ from Kosinski’s algorithm and how does it work?
It is difficult to definitively state what caused changes in the American swing states. In any case, I am glad to see that media attention has shifted to more important topics such as the likelihood that powerful organizations and individuals in Trump’s administration use techniques to manipulate voters based on big data.
– Michal Kosinski developed models that are integral to the Apply Magic Sauce (AMS) algorithm, and we continue to collaborate with him. The AMS project is unique. No other personality prediction service is based on a larger and richer database – for the project, we received data from six million social media users. AMS uses likes and statuses from Facebook and other digital footprints to generate psychological predictions about an individual in half a second. The algorithm compares the digital footprints of millions of users who have completed psychometric tests and agreed to share their social media data with us. By recognizing patterns in digital footprints, the algorithm selects subtle signals that it considers indicative of certain psychological characteristics.
- In which areas can psychometric techniques contribute the most?
– Psychometrics has a long history and a promising future, especially in the fields of education, career, and clinical testing. Driven by our research, psychometrics has recently entered the online world and brought new opportunities. Computerized testing has revolutionized examinations in the field of education and made significant progress in terms of fairness, objectivity, and meritocracy. Tests can now be shorter, more precise, and, if conducted properly, harder to cheat on. The business and healthcare sectors lag behind in this regard. I notice an excessive reliance on unscientific tools, which may only indicate the obsolescence of the testing market. Marketing and sales efforts to categorize employees into certain types, colors, or other arbitrary categories have replaced common sense in questioning what characteristics make a good employee. Recruiters have very poor insight into the quality of psychometric tests, their reliability, and validity.
- How can psychometric techniques contribute to the business world?
– Psychometrics applied in big data can inspire systems that adapt to our unique desires and psychological preferences in real-time. Unlike most actuarial systems used in insurance or banking, they can explain their decisions and connect decision-making with human traits, rather than data correlations that are impossible to interpret.
- How accurate are these predictions and is it risky to rely entirely on them?
Facebook earns an average of less than two dollars per month per user from advertising. Therefore, the do-not-track option could be monetized. I would gladly subscribe to it and pay for my privacy. However, we must be careful that privacy is not elitized.
– The personality prediction algorithm can surpass the accuracy of human judgment in determining results on personality tests. Based on just 10 likes, the computer assessment exceeds the judgment of colleagues, based on 70 likes it surpasses the judgments of friends, and with more than 300 likes it exceeds the assessment ability of a spouse. It is always risky to rely solely on one technology, but this is not much different from relying on just one selection test or just one interview with an employer. Most companies use several tools and compare their relative strengths and weaknesses. Personality prediction based on digital footprints is fast, cheap, and accurate for certain individuals, but it should not be used as the sole source of data when the stakes are high. Nevertheless, companies regularly rely on much less when hiring, firing, or choosing a new strategic direction.
- During the American presidential election campaign, candidate Donald Trump hired Cambridge Analytica, which advised him using big data. The media has been (and still is) full of reports about how big data played a crucial role in Trump’s victory. Is psychology of big data responsible for his victory?
– I am not a political expert, but I have not seen any evidence that big data analysis, micro-targeting, or any other technique allegedly used by Cambridge Analytica influenced Trump’s victory. I believe that candidates win elections, not big data. Since it is not transparent how the campaign was conducted, it is difficult to definitively state what caused changes in the swing states. It has been mentioned in the works of Plato, Aristotle, and even some earlier writings that messages appealing to emotions have greater persuasive power than those directed at reason. (Interestingly, some employees of Cambridge Analytica in their recent statements claim that they never used psychographic techniques during the campaign.) In any case, I am glad to see that media attention has shifted to more important topics such as the likelihood that powerful organizations and individuals in Trump’s administration use techniques to manipulate voters based on big data.
- How are big data and predictive technologies being used in election campaigns currently underway in Germany and France, and were they used in the Brexit campaign?
– I am sorry, I cannot answer those questions. I have no information on whether big data techniques were used in European elections or in Brexit. There were some indications in media releases from Cambridge Analytica that they were seeking clients in Europe, but they are very secretive about whom they collaborate with. Although I am sure they participated in the Leave.EU campaign (according to research by the Observer and the Guardian), I do not know exactly what they did. I think it would be naive to believe that they are not collaborating or at least trying to collaborate with right-wing clients in Europe like the National Front or AfD.
- Research you conducted in 2016 showed that 71 percent of people believe that companies with access to their personal data do not handle that data ethically. Will the General Data Protection Regulation (GDPR) improve the situation?
I have no information on whether big data techniques were used in European elections or in Brexit. It would be naive to think that Cambridge Analytica is not collaborating or at least trying to collaborate with right-wing clients in Europe like the National Front or AfD.
– It is true that many companies have a lot of work to do to convince the public that they are responsibly using their personal data. I was surprised when I saw that result in our research ‘Trust and Predictive Technologies Report 2016’ because I am aware of how serious the financial consequences, and even reputational consequences, of data misuse can be. Other research, such as that conducted by Accenture, shows that most companies are still not ready for the provisions of the GDPR. Therefore, I expect that in the first few months after its implementation, there will be many violations and strict penalties will be imposed. Although legislation offers effective protection in data collection, most laws, including the GDPR, do not recognize how non-personal, seemingly innocuous data can be combined with other databases and processed in a way that makes it easy to identify an ‘anonymous’ individual. There are many real-world examples where databases have ceased to be anonymous, such as the publication of viewing history on Netflix or taxi rides in New York. It is always possible for data to leak.
- Is it moral to use personal data?
– Digital citizens also have their share of responsibility in this. With mobile phones and always-on services that track us throughout the day, we must accept that most of the data we would not want companies to use is already shared in one form or another. A click on Google, a picture on Instagram, a physical activity tracking app – and the data is there. Therefore, the point is not to control how data is collected, but rather how it is later used and how to ensure that individuals are rewarded for sharing their data. A positive example of this is CitizenMe, a London-based market research company with which the Centre collaborates. It allows users of its app to exchange personal data with brands, answer their questions, and so on, and in return, they are rewarded with insights into their psychological profile or monetary compensation.
- How to encourage citizens to rebel against the collection and trading of personal data without their consent or appropriate compensation?
Based on just 10 likes, the computer assessment in just half a second exceeds the judgment of colleagues, based on 70 likes it surpasses the judgments of friends, and with more than 300 likes it exceeds the assessment ability of a spouse.
– We believe that one of the first steps is to teach citizens that their data has value. Most people understand that health insurance can change the premium payment if it knows what they eat, how much they exercise, and so on. However, few realize that clicks and likes can create a very sensitive psychological profile that reveals their desires and sexual orientation. Initiatives like Apply Magic Sauce and other open-source tools we publish are key in showing what can be done with data. This helps engage all stakeholders, including citizens, in an open discussion about how predictive technologies should be used. I do not believe that digital withdrawal is the right goal. We want to encourage companies to design better services that understand our personalities and better meet our needs. I can imagine a future where it is more profitable to protect than to violate privacy and where competitive advantage is not the amount of data, but the efficiency in turning that data into benefits for the user.
- Companies offer users free services and then profit from the data they collect from them. Have we become a product? Should we start paying providers of ‘free’ services to preserve our privacy?
– If you do not pay in cash, you will pay with privacy. I do not agree with some privacy advocates who believe that companies like Facebook and Google should continue to provide their services for free simply because we have become accustomed to them in our daily lives. Facebook earns from advertising and therefore must track its users. On average, Facebook earns less than two dollars per month per user from advertising. Therefore, the do-not-track option could be monetized. I would gladly subscribe to it and pay for my privacy. However, we must be careful that privacy is not elitized, meaning that it does not happen that only those who can afford it can have it. It remains one of the fundamental rights in European legislation, and we should persist in creating innovative business models that promote equality in terms of personal data privacy.
- When it comes to big data, what does the near future hold for us?
– Artificial intelligence and the process of robotization will play an important role in the future. Small and large organizations have ridden this wave despite not fully understanding what the ultimate outcome will be. Technological advancement has made data storage and processing cheaper than ever, and thus it is less likely that companies will adhere to the principle of deleting all data that is not used. According to Veritas research, most of the data that companies retain is completely useless; only 14 percent of stored data is essential for business. My interest in the future is to explore the possibilities of combining digital footprint predictions with blockchain systems and creating a personal data exchange system that allows users real-time access to that data and compensation for it. A hybrid solution of this kind could be another piece of the puzzle in our endeavor to personalize the internet.
The interview was originally published in the 600th issue of Lider on March 31, 2017.
