186 - Guest: Michal Kosinski, Professor of Psychology, part 1
Jan 8, 2024
auto_awesome
Guest Michal Kosinski, Professor of Psychology, discusses the Cambridge Analytica scandal, using Facebook likes to predict traits, the concept of privacy today, challenges and benefits of personal data collection, and the intersection of privacy, AI, and creativity.
Machine learning models can accurately predict people's intimate traits from their digital footprints, raising ethical concerns and potential dangers.
Companies like Facebook and Google can create highly targeted advertisements using algorithms to exploit individuals' vulnerabilities and influence their actions.
Deep dives
Cambridge Analytica Scandal and Predictive Models
In this podcast episode, Michal Koshinsky discusses his early research on privacy that preceded the Cambridge Analytica scandal. Through his research, Koshinsky demonstrated that personal traits and attributes can be predicted from individuals' digital footprints, such as Facebook likes. Despite initial skepticism and dismissal from academia, his study showed that machine learning models can accurately predict people's intimate traits, including political orientation and personality. Koshinsky also highlights the ethical concerns and potential dangers associated with the unauthorized acquisition of personal data and the use of these predictive models.
Facebook's Role and Data Access
Koshinsky sheds light on Facebook's involvement in predictive modeling and the access the platform had to users' data. He reveals that back in 2012, Facebook published a patent disclosing a technology that could accurately determine people's intimate traits from their digital footprints. This raised concerns among researchers and policymakers about privacy violations, as this predictive model could reveal information about users' personality, intelligence, and political orientation without their knowledge. Koshinsky emphasizes that while Cambridge Analytica garnered attention for exploiting this technology, larger players like Facebook, Google, and Amazon have even greater access to user data, making the potential risks and implications of predictive modeling even more significant.
Personalized Advertising and Manipulation
The podcast episode explores the power and implications of personalized advertising enabled by predictive modeling. It highlights how companies like Facebook and Google can create highly targeted advertisements by using algorithms to analyze users' digital behavior, interests, and demographic information. This highly personalized advertising can be tailored to specific psychodemographic characteristics, such as political orientation, age, location, and more. Koshinsky underscores the potential manipulation involved in this form of advertising, as companies can use psychological science to exploit individuals' vulnerabilities and influence their actions and opinions. He raises concerns about the lack of societal input in regulating the rules and boundaries of advertising on these platforms.
The Post-Privacy Era and Technological Progress
Koshinsky discusses the prevailing post-privacy world we live in today, where motivated third parties can access vast amounts of personal data to know more about individuals than they know about themselves. He acknowledges the convenience and benefits of sharing data, which allow technology platforms to provide personalized recommendations and services. However, he also underscores the risks posed by the loss of privacy and the potential misuse of personal information for various purposes, including consumerism, political manipulation, and societal control. Ultimately, Koshinsky highlights the need for a balanced approach to data privacy, where societal institutions participate in the regulation and governance of data usage rather than leaving it solely in the hands of profit-driven companies.
The worlds of academia and political upheaval meet in my guest Michal Kosinski, who was behind the first press article warning against Cambridge Analytica, which was at the heart of a scandal involving the unauthorized acquisition of personal data from millions of Facebook users and impacting the 2016 Brexit and US Presidential election votes through the use of AI to microtarget people through modeling their preferences.
Michal also co-authored Modern Psychometrics, a popular textbook, and has published over 90 peer-reviewed papers in prominent journals such as Proceedings of the National Academy of Sciences (PNAS), Nature Scientific Reports and others that have been cited over 18,000 times. Michal has a PhD in psychology from the University of Cambridge, as well as master’s degrees in psychometrics and social psychology, positioning him to speak to us with authority about how AI has and may shape the beliefs and behaviors of people en masse.
In this first part of the interview, we delve into just that, plus the role of social media, and Michal's take on what privacy means today.
All this plus our usual look at today's AI headlines.