3Summary (Дифференциация учащихся по академической успеваемости в социальной сети)

NATIONAL RESEARCH UNIVERSITY HIGHER SCHOOL OF ECONOMICSINSTITUTE OF EDUCATIONIvan SmirnovStudents’ differentiation by academic performanceon a social networking siteSummary of the PhD thesisTranslated from RussianAcademic supervisor:Prof. Katerina PolivanovaDoctor of ScienceMoscow — 20182The thesis was prepared at the National Research University Higher School of Economics.Publications:1.

Smirnov I., Sivak E., Kozmina Y. In Search of Lost Profiles: The Reliability of VKontakteData and Its Importance for Educational Research // Educational Studies. 2016. №4. С. 106122.2. Polivanova K., Smirnov I. What’s in My Profile: VKontakte Data as a Tool for Studying theInterests of Modern Teenagers.

// Educational Studies. 2017. №2. C. 134-151.3. Smirnov I., Thurner S. Formation of homophily in academic performance: Students changetheir friends rather than performance. // PloS one. Vol. 12. №. 8, e0183473.URL: https://doi.org/10.1371/journal.pone.01834734.

Smirnov I. Predicting PISA scores from students’ digital traces // Proceedings of the TwelfthInternational Conference on Web and Social Media. American Association for ArtificialIntelligence (AAAI) Press, 2018. P. 360-364Selected Conferences:— RAHER — VII International Conference of Higher Education Researchers, Moscow, Russia,2016.— XVIII April International Academic Conference on Economic and Social Development,Moscow, Russia, 2017.— NetSci-X – International Conference on Network Science, Tel Aviv, Israel, 2017.— IC2S2 — International Conference on Computational Social Science, Cologne, Germany, 2017.— SocInfo — International Conference on Social Informatics, Oxford, United Kingdom, 2017.— RAHER — VIII International Conference of Higher Education Researchers, Moscow, Russia,2017.— XIX April International Academic Conference on Economic and Social Development, Moscow,Russia, 2018.— ICWSM — International AAAI Conference on Web and Social Media, Stanford, USA, 2018.3INTRODUCTIONScientific tools have always played an important role in the scientific progress.

Theinvention of the telescope opened a new vast world of celestial bodies. Observations of theplanetary motion laid the foundation of the classical mechanics and gave the rise to physics as ascience in the modern sense. The invention of the microscope opened a new vast world ofmicroorganisms and laid the foundation of the modern biology and medicine. Today, we observecomparable changes in the social science [1–3]. In place of optical instruments are methods ofcollecting and analyzing digital traces, vast amounts of data made available by the recentdevelopment of information technology.Digital traces allows researchers to see what was previously hidden from them. For instance,educational researchers have long been studying social relationships between students [4–9].However, for a long time, the only method for collecting the data on students’ social ties wassurvey. Surveys only allow reconstructing the social network of a small closed group of people.

Asa result, researchers well understand the social networks of students within one school but knowalmost nothing about interschool social ties. In contrast, Facebook has information about friendshipties of two billion people1. Information about social ties need not necessarily be explicitly indicated,as in the case of friendship on a social networking site. It can effectively be reconstructed fromphone calls logs [10] or from information about physical proximity from mobile sensors [11]. Thedirect access to information about human behavior also resolves the problem of censoredness andsocial acceptability characteristic of survey answers [12].Another advantage of digital traces is a significant reduction in the cost of research.

Thelargest international study on education PISA covers about 500,000 students [13]. Its organizationrequires a large amount of resources and coordinated effort from dozens of teams in differentcountries. At the national level, studies with several thousand participants are considered as large.The thesis could be successfully completed with hundreds or even dozens of respondents. Datafrom social networking sites allows individual researchers to work with much larger samples up totens of millions of participants [14–16]. Data need not necessarily be accessible directly. It waspreviously shown that various socio-demographic characteristics such as ethnicity, gender orincome level could be effectively predicted from tweets [17], profile images [18], user posts [19], orphotos of neighborhoods [20].Traditionally, social sciences make an emphasis on theories and their ability to explainhuman behavior while ignoring their predictive power which is often much inferior to natural1https://newsroom.fb.com/company-info/4sciences [21, 22].

This significantly limits the possibility of practical application of suchknowledge. One of the reasons for the weak predictive power lies in the fact that social systems andthe laws that govern them are too complex to be described by simple variables [23]. For instance, avariable with a few discrete values cannot describe the wide variety of personal characteristics andhuman behavior even in theory.

Today, thanks to big data, one could construct new variables withmuch greater predictive power. For instance, it has been shown that information about the likes on asocial networking site allows to predict various personal traits from sexual preferences tointelligence and parents’ divorce [24]. While individual likes do not have much predictive power,together they form a unique digital trace that contains a significant amount of information about anindividual.Thus, the use of digital traces for studying the human behavior is one of the most promisingand actively developing areas in the social science.

This approach has a great potentiation in theeducation science as well. In the Russian context, the most promising source of data is the socialnetworking site VK, that is already been used in educational research [25–27]. VK is of particularinterest not only because it provides an access to a vast amount of data but also because of its role inthe lives of modern students.Today, young people are almost always online [28, 29] and spend most of their online timeon SNSs [30]. Recently SNSs have become the main source of traffic to other resources of theInternet [31]. In Russia, the largest SNS VKontakte (VK) is named as the main source ofinformation about the country and the world by 70.3% of young people, more than any other sourceof information [32].

This SNS is also more trusted by young people than traditional media [32].Thus, VKontakte largely determines the information environment of modern Russian students. Thatis why it is particularly important to study the nature of this environment and to understand whetherit is different for different groups of students.This work is devoted to the development of methods for mining and analyzing VK data inthe context of educational research.

These methods are applied to study differentiation of studentsby academic performance. The need to develop the methods is due to the fact that little is knownabout the reliability of the data and the potential biases of VK samples and that analysis of bigmultidimensional data requires special approaches. The differentiation of students on a socialnetworking site is of particular importance in the context of digital inequality [33–35].5Theoretical frameworkThe authors of the concept of digital inequality note that one generalization that emergesfrom research on inequality in access to information about cultural resources is the differentiationprinciple.

This principle means that at the beginning — when access to a particular resource islimited — the use of this resource remains relatively undifferentiated. When the resource becomesmore accessible, its use becomes more differentiated. It is explained by the fact that more-privilegedgroups find a way to use the resource to extract more from it than less-privileged groups [33].The typical example here is education. When access to secondary education became almostuniversal, most children from the middle and upper classes started to attend universities. At thatpoint, university degrees became the main differentiating characteristic. When higher educationbecame widespread, two groups of universities emerged: mass universities and selectiveuniversities, and the main differentiating characteristic shifted from having any higher educationdegree to graduation from the prestigious university [33].The same principle applies to digital resources.

In theory, the global network provides allstudents with equal access to virtually unlimited knowledge. However, as early as the mid-1990s,researchers became concerned about the digital divide between those who have or do not haveaccess to computers (later to the Internet) [36]. When the Internet became ubiquitous, the concernwas about the second digital gap related to the gap in the level of information literacy and the abilityto use all the opportunities that are provided by the Internet. Today, the main focus is shifted to abroader concept of digital inequality, which is understood as the difference in the Internet use (orthe use of any other digital technology) between groups of people who have formally equal accessto it [33].The existing research [37, 38] suggests that there are significant differences in the Internetuse between users with different levels of formal education.