I am an Assistant Professor at the Business School of the University of Mannheim, Germany. I am using AI methods to study biases in social systems and borrow methods from social sciences to understand biases of AI.
Previously, I founded and lead a Computational Social Science Group at the Higher School of Economics, Moscow. We were using big data and machine learning to better understand human behavior and complex social phenomena. Our main focus was on inequality in education and the psychological well-being of students.
Our research was supported by the Russian Science Foundation, presented at IC2S2 and ICWSM, and published in Proceedings of the National Academy of Sciences, EPJ Data Science, and Royal Society Open Science. Our work was also covered by leading Russian and international media including MIT Technology Review, The Times, and Nature.
To further promote computational social science in the country, I have developed and taught a course Introduction to Computational Social Science — the first of its kind in Russia — and co-organized the Summer Institute in Computational Social Science at HSE.
I then moved to Germany where I joined Markus Strohmaier's group first at RWTH Aachen and then at the University of Mannheim.
I have also been involved in social entrepreneurship, helping to launch Teach for Russia RU and Sci.STEPS RU.
In my free time, I am working on an open online course Introduction to Computational Social Science RU.
While there are many studies on the friendship between students, most of them focus on students from a single educational institution, i.e. study friendship ties within one school or one university. As a result, little is known about social connections between students from different schools. In this paper, I have used digital traces to investigate interschool friendship on a scale of the whole city. I have analyzed data on 37,000 students from 590 schools and their friendship links on VK and have found that students from similar performing schools tend to become online friends. One might assume that this is a trivial consequence of the geographical segregation of schools. However, by adding data on school locations and apartment prices, I was able to show that segregation in the digital space is in fact much stronger than geographical segregation.
In this paper, I have built a model to predict the academic performance of students from their posts on social media. I have combined unsupervised learning of word embeddings on a large corpus of social media posts with a supervised model trained on data from a nationally representative sample of young adults. This data set contains the academic performance of students measured by a standardized test as well as information on their public activity on social media. I have used a continuous-vocabulary approach that allowed achieving high accuracy using a relatively small training data set. It also allows computing interpretable scores for millions of words that are fun to explore!
Parents’ preference for sons is a well-known phenomenon that manifests in various forms from sex-selective abortions to higher investments in sons. In this paper, we used public posts made by 635,665 users on a popular Russian social networking site, to investigate public mentions of daughters and sons on social media. We find that both men and women mention sons more often than daughters in their posts. We also find that posts featuring sons receive more “likes” on average. Our results indicate that girls are underrepresented in parents’ digital narratives about their children. Previous studies have shown female characters are underrepresented in children’s books, textbooks, movies, and on Wikipedia. Gender imbalance in public posts may send yet another message that girls are less important and interesting than boys and deserve less attention, thus presenting an invisible obstacle to gender equality.