Lab members defended their final qualification papers
This season, three members of the International Laboratory for Applied Network Research — research assistants Polina Ermolaeva, Mikhail Meshkov, and Anton Sizov — defended their final qualification theses. All received the grade "Excellent."
Master's program: "Data Analytics for Government and Society". Academic supervisor: Valeria Ivanyushina
"Selection and influence of peers on adolescent sexual debut in Russian vocational colleges"
Peer influence on adolescents' visible risk behaviors — smoking, drinking — is well documented. Effects on private behavior, particularly sexual debut, are far less understood. Polina tested both mechanisms, influence and selection, using longitudinal friendship network data from vocational colleges in St. Petersburg (20 groups, three waves), applying stochastic actor-oriented modeling.
Result: both mechanisms are at work. Adolescents are more likely to start their sexual life if they associate with sexually active peers (influence), and more likely to befriend peers with a similar status (selection). At the same time, sexually active adolescents received fewer friendship nominations on average — regardless of gender, which does not support the sexual double standard hypothesis.
"I worked on this thesis with great interest throughout my master's studies — learning new methods and theories, presenting preliminary results at conferences. The defense went smoothly, and I'm happy with my grade."
Mikhail is a student in our master's program "Data Analytics and Social Statistics". Academic supervisor: Daria Maltseva
"Comparative analysis of clustering results for keyword co-occurrence networks"
Mikhail compared the robustness of three community detection algorithms — Leiden, Infomap, and Walktrap — on keyword co-occurrence networks built from 725 publications on AI applications in clinical research. Keywords were drawn from OpenAlex (Keywords Plus) and PubMed (MeSH and author keywords). Robustness was tested through 30 stochastic runs and by removing 5% of edges and 5% of publications.
Conclusion: Infomap fails to detect communities in dense Keywords Plus networks but proves robust on the sparser author-keyword network; Leiden produces balanced partitions but loses robustness on author keywords; Walktrap tends toward excessive fragmentation. Algorithm choice should account for the keyword source and network topology — and be documented at every stage.
Mikhail's advice to future students in the program: choose your thesis topic from your first year and build your term papers so that each becomes a stage of the thesis — this saves effort and ensures a quality result. "Tried and tested on myself."
Bachelor's program: Sociology. Academic supervisor: Maria Rodionova
"'Chilling effects' of state surveillance: determinants of Russians' political participation, self-censorship practices, and digital resistance"
Earlier research failed to establish a causal link between perceived state surveillance and digital political participation, due to methodological limitations. Anton proposed sociological foundations for chilling-effects theory and an analytical framework of the contexts in which awareness of online surveillance arises. Using data from a three-wave panel online survey (August 2025, N=1554; December 2025, N=1280; April 2026, N=1419), the effect was confirmed: perceived surveillance reduces political deliberation on social media and consumption of political content on blocked platforms. The effect is mitigated by privacy-protection practices, involvement in NGOs, and larger ego-network size.
Congratulations to our colleagues on their successful defense, and best wishes for continued success on their path ahead!
