My Research

I employ big data and computational models to address pressing issues, including privacy risks, psychometrics, online mass persuasion, and psychological profiling. In pursuing my research, I have published over 70 papers, which have been cited over 11,000 times. The following subsections summarize my work and its implications. 

A list of all of my papers can be found in my CV.


Birds of a feather do flock together: behavior-based personality assessment method reveals personality similarity among couples and friends (2017)

Youyou, Schwartz, Stillwell, & Kosinski; Psychological Science (Kosinski is the second & senior author with a student co-author)

Computer-based personality judgments are more accurate than those made by humans (2015)

Youyou, Kosinski, & Stillwell; Proceedings of the National Academy of Sciences (Kosinski shares the first authorship with a student co-author)


And also …

Links Between Facial Features and Psychological Traits

The theoretical foundations of these links are well established. Faces shape others’ perceptions and expectations that, in turn, shape our behavior and psychological traits. Psychological traits and facial features are both influenced by many common factors, such as hormones and genes. Finally, psychological traits influence facial appearance; for example, attitudes affect group memberships that, in turn, influence facial grooming style. 

Spouses’ faces are similar but do not become more similar with time (2020)

Tea-makorn & Kosinski; Scientific Reports by Nature

Both old wives’ tales and psychological literature posit that spouses’ faces become more similar over time. Surprisingly, this widely accepted belief is supported by a single 1987 study of 12 couples, whose results have never been replicated. We validate it in a large sample of 517 couples using two independent methods of estimating facial similarity (human judgment and a state-of-the-art facial recognition algorithm). Our results show that while spouses’ faces tend to be similar at the beginning of marriage, they do not converge over time. 


Facial Width-to-Height Ratio Does Not Predict Self-Reported Behavioral Tendencies (2017)

Kosinski; Psychological Science

Many studies have linked facial width-to-height ratio (fWHR) with various antisocial and violent behavioral tendencies. However, those studies have predominantly been laboratory-based and low-powered. This paper showed that the links between fWHR and psychological traits were weak, and that their patterns were inconsistent with the previous findings and theoretical assumptions advanced in the fWHR literature. 

And also …


Facial recognition technology can expose political orientation from naturalistic facial images (2021)

Kosinski; Scientific Reports by Nature

Ubiquitous facial recognition technology can expose individuals’ political orientation, as faces of liberals and conservatives consistently differ. A facial recognition algorithm was applied to naturalistic images of over 1 million individuals to predict their political orientation by comparing their similarity to faces of liberal and conservative others. Political orientation was correctly classified in 72% of liberal–conservative face pairs, remarkably better than chance (50%), human accuracy (55%), or one afforded by a 100-item personality questionnaire (66%). Accuracy was similar across countries (the U.S., Canada, and the UK), environments (Facebook and dating websites), and when comparing faces across samples. Accuracy remained high (69%) even when controlling for age, gender, and ethnicity. Given the widespread use of facial recognition, our findings have critical implications for the protection of privacy and civil liberties. 


Deep neural networks are more accurate than humans at detecting sexual orientation from facial images (2018)

Wang & Kosinski; Journal of Personality and Social Psychology (JPSP)

After discovering that governments and corporations had been using facial analysis algorithms to detect intimate traits, we decided to examine the  privacy threats posed by such algorithms in the context of one of the most sensitive personal characteristics: sexual orientation. 

We showed that an algorithm can detect sexual orientation from facial images with an accuracy comparable with that of mammograms or modern diagnostic tools for Parkinson's disease. The differences between the faces of gay and straight people were consistent with widely accepted theories explaining the origins of sexual orientation.

The study has been successfully replicated by two research teams using independently collected data: 


Psychological targeting as an effective approach to digital mass persuasion (2017)

Matz, Kosinski, Nave, & Stillwell; Proceedings of the National Academy of Sciences (PNAS) (Kosinski is the second author & senior co-author)

This paper illustrated the feasibility and effectiveness of the mass persuasion techniques deployed by companies such as Cambridge Analytica and exposed the risks inherent to behavioral targeting, the principal source of income for many tech giants. 


See also:

Private traits and attributes are predictable from digital records of human behavior (2013)

Kosinski, Stillwell, & Graepel; Proceedings of the National Academy of Sciences (PNAS)

Upon realizing the extent to which companies and governments use people's digital footprints to extract their intimate traits, we decided to see how accurate such predictions are using one of the most pervasive and easily accessible digital footprints, Facebook Likes.  Our results showed that they can be used to automatically and accurately predict a range of sensitive traits, including sexual orientation, ethnicity, political views, personality, intelligence, use of addictive substances, parental separation, etc. 

This paper and accompanying op-ed in Financial Times called for, and triggered, an immediate tightening of Facebook’s privacy policies: Two weeks after its publication, Facebook switched off public access to its users’ Likes (until then, they were by default visible to anyone on the internet). This greatly limited exposure to privacy threats for millions of users and reduced the damage caused by Cambridge Analytica. 


And also …


Modern Psychometrics: The Science of Psychological Assessment 4th Edition (2020)

Rust, Kosinski, & Stillwell; Routledge.

Computational Social Science (chapter) 

Kosinski; in Advanced Social Psychology: The State of Science 2nd Edition (2019) by Finkel & Baumeister; Oxford University Press.

This volume -- an update to the original, 2010 edition -- provides a graduate-level introduction to social psychology. The target audience consists of first-year graduate students (MA or PhD) in social psychology and related disciplines (marketing, organizational behavior, etc.), although it is also appropriate for upper-level undergraduate courses. The authors are world-renowned leaders on their topic, and they have written state-of-the-art overviews of the discipline's major research domains. The chapters are not only scientifically rigorous, but also accessible and engaging. They convey the joy, excitement, and promise of scientific investigations into human sociality. 

Mining Big Data to Extract Patterns and Predict Real-Life Outcomes (2016)

Kosinski, Wang, Lakkaraju, & Leskovec; Psychological Methods.

Facebook as a Research Tool for the Social Sciences: Opportunities, Challenges, Ethical Considerations, and Practical Guidelines (2015)

Kosinski, Matz, Gosling, Popov, & Stillwell; American Psychologist

And also …


Psychology of Music

Other Journal Publications