Conference abstracts


Keynote talk

4 July, 15:30, Conference hall

Tom Snijders, University of Groningen

Dynamics of Multilevel and Multivariate Networks

The term ‘multilevel’ may be understood in a variety of ways. One of my favourite meanings is to let it refer to social science models involving multiple types of actors. Examples are actors and ties, or actors and events, or individuals and organizations, or people and concepts.

Network analysis is basically multilevel just by itself, because it involves actors and ties. This leads to interesting complexities in treating the effects of actor attributes on networks. Some of these often are overlooked. Network analysts will automatically think of homophily, but perhaps stop there. However, for numerically valued actor attributes there are additional mechanisms such as aspiration to select network partners with high and desirable values; a conformist tendency to relate to others who are ‘quite normal’; and the sociability that may be higher for actors having high values. A way will be presented to express combinations of such mechanisms in statistical network models.

For multivariate networks, the multilevel nature of networks is evident even more naturally. Three basic ways in which one network can influence another one are dyadic entrainment, at the level of ties; popularity and activity effects, at the level of actors; and cross-network transitive closure, together with other algebraic effects at triadic or higher levels.

Multilevel networks can be defined as structures with several node sets, with several different networks of which the meaning differs, depending on the node sets they connect. A basic example is the combination of a one-mode and a two-mode network, e.g., a one-mode friendship network and a two-mode activity network; or a network of cooperation between actors and a network of the use of concepts or tools by these actors. Here both of the issues raised above appear, and lead to issues such as the definition of homophily for a two-mode network, and mixed transitive closure for one-mode – two-mode network combinations.

All this will be discussed in the context of statistical dynamic network modelling. Co-evolution of multiple networks is a fruitful approach here, offering a framework to ask questions such as ‘do we collaborate because we use the same concepts, or do we use the same concepts because we collaborate?’. A downside of all this interesting complexity is that it leads to models having many parameters to be estimated…

Socio-Semantic Networks. Part 1.


4 July, 17:00, Room 140

Camille Roth, Sciences Po, Paris and Centre Marc Bloch, Berlin

Articulating the local and global socio-informational dynamics (case studies from German-, French-, and English-speaking digital public spaces)

Socio-technical systems involve actors who create and process knowledge, exchange information and create ties between ideas in a distributed and networked manner: online social networks, blogging and micro-blogging platforms, scientific communities are, among others, examples of such systems. The state-of-the-art in this regard focuses on two main issues which are generally addressed in a relatively independent manner: the description of content dynamics and the study of interaction network characteristics and evolution. Further, many approaches focus either on the ego-centered level (describing for instance the variety of ego's immediate neighborhood) or on the global level (e.g., by describing the configuration of clusters). After reviewing the relevant state of the art, we will propose directions as to how to articulate both the actor and the group levels, and/or the interactional and informational levels. We will specifically present a variety of examples related to online communities, with some case studies related to the German-, French- and English-speaking digital public spaces.

Vladimir Batagelj, Institute of Mathematics, Physics and Mechanics, Ljubljana

Fractional bibliographic coupling and co-citation

In a network of works (papers, books, etc.) the citation relation Ci means: p Ci q ≡ work p cites work q. Using the network multiplication * (Batagelj and Cerinšek, 2013) the bibliographic coupling (Kessler, 1963) network biCo can be determined as biCo = Ci*t(Ci), where t(N) is a reverse/transpose of the network N. biCo[p,q] = # of works cited by both works p and q = |Ci(p)∩Ci(q)|. This suggests some content communality between p and q. It is thought that sharing more cited works increases the likelihood of them sharing a content. In searching for works with similar content we noticed that the works citing many works - especially review works - are overrated. To neutralize their distorting impact we define a fractional bibliographic coupling, using normalized measures. We start with a network biC = n(Ci)*t(Ci), where n(Ci) = D*Ci and D = diag(1/max(1,outdeg(p))). For a nonempty Ci(p) it holds biC[p,q] = |Ci(p)∩Ci(q)|/|Ci(p)| and biC[p,q]∈[0,1]. biC[p,q] is the proportion of its references the work p shares with the work q. A fractional bibliographic coupling measure can be defined as some mean (max, min, arithmetic, geometric, harmonic, Jaccard, etc.) of biC[p,q] and biC[q,p]. We will discuss different measures we obtain, their properties, relations among them, and their computation.Two works are co-cited if there is a third work that cites both works (Marshakova, 1973; Small, 1973). The co-citation network coCi can be determined as coCi = t(Ci)*Ci. coCi[p,q] = # of works citing both works p and q = |t(Ci)(p)∩t(Ci)(q)|. coCi[p,q] = coCi[q,p]. The co-citation is a kind of “dual” (replacing Ci with t(Ci) ) notion to bibliographic coupling.For illustration we present some results of applying the proposed measures on some (large) bibliographic networks. The analyses were performed using Pajek – a program for analysis and visualization of large networks (De Nooy et al., 2011; Batagelj et al., 2014).

Jan Rasmus Riebling, University Wuppertal; Bamberg Raphael Heiko Heiberger, University of Bremen

Semantic Bridges. Measuring Thematic Similarity in Economists' Collaboration by Combining Topic and Temporal Exponential Random Graph Modeling

Textual data of all sorts is increasingly popular in network research and used to retrieve information on social and semantic relations. Yet, positions and properties of the creators of texts are almost never incorporated in modeling the semantic space, social science research has mostly investigated the phenomena of social or semantic structure separately. The structural positions and properties of the creators of texts are very seldom incorporated in models of the semantic space.

Similar assessments can be made regarding approaches stemming from natural language processing, e.g. most variants of topic models omit variables linked to the documents though they often comprise meaningful information on social structures.

To combine social and semantic relations we employ on the one hand network and structural data from collaborations of economists and, on the other hand, textual data from their abstracts. The sample consists of all articles from the top-100 journals in economics between 1990 and 2015. Utilizing thematic orientations of authors as covariates in "Temporal Exponential Random Graph Models" allows us to investigate the semantic homophily of co-authors net of other potential influential factors like prestige, institutional affiliations or previous collaboration. In doing so we amplify the rich literature of team science and emphasize the role which semantic bridges play in economic research and provide a model that can be easily adopted to other fields of knowledge.

Qualitative Network Analysis. Part 1


4 July, 17:00, Room 213

Iulia Mihalache, Université du Québec en Outaouais

Human and Nonhuman: Translation Networks

Translators and other language professionals organize themselves in networks to spread knowledge about translation processes and practices, share resources, run projects, receive training or act in humanitarian situations. Sociological theories of translation (Chesterman 2006; Buzelin 2013; Wolf & Fukari 2007) regard translation as an interactive social event where meaning is created or transcreated in a new sociocultural, ideological context. Translation contributes to spreading ideas, which can become contagious across cultures (Sperber 1996), in the same way viruses, worms, and other software objects contamine networks while at the same time articulating their social connections to political and cultural discourses (Parikka 2016; Sampson 2012). In the field of translation studies, the actor-network theory (ANT) has been used to investigate translation practices such as collaboration, mediation or crowdsourced translation, with “translation” understood not as a linguistic transfer, but as a metaphor able “to unveil the strategies actors use to enrol others and fulfil their objectives.” (Buzelin 2013: 189). These strategies could also be seen as “stratagems”, revealing tensions and asymmetries in networks (Estrada 2016), a politics of transmission and communication, a work of manipulation and hijacking (Fuller & Goffey 2012) where technologies extend into every aspect of human life and remain mysterious, almost magical (like neural networks in machine translation). Networks are alive, spawning connections between human and nonhuman agents. Translation networks themselves can be either human or nonhuman (machine-based, artificial), but acting as social, communicating beings, with their own rhetorical, organizational capital. The human and nonhuman are entangled. This presentation will look at both human and nonhuman translation networks, focussing on the politics of transmission and communication in the network.

Kay Junge, Bielefeld University

The tools of exchange on the Kula

How the Kula Ring´s circular structure might explain the specific type of ceremonial and economic exchange and, inversely, how the particular items exchanged might have given rise to its particular structure are still questions of central concern in current research. This paper divides into two sections, one on the topology of the network, the other on the tools of trade used to make the Kula go round. I will first review the three main theoretical attempts to model the Kula´s ring-like structure. Ziegler (1990) focuses on the evolution of trust by conceiving of the local interactions along the ring in terms of a repeated prisoners dilemma; Landa and Grofman (1983) explaining the structure of the network in terms of efficient transportation, and the more recent attempts by Skyrms (2010) and Goyal (2007) explaining the evolution of particular network structures by their differences in facilitating information transfer. The tools of trade, i.e. the necklaces and arm-shells used for the ceremonial establishment of the real trade going on around the Kula have proofed more resistant to theoretical reconstructions. Neither the notion of primitive money nor that of magic turned out to be very helpful. I hope to shed some new light on the use of these items by comparing them with the symbola and tessera instituting foreign trade relations in Ancient Greek, with the tally sticks used in medieval England and the bills of exchange and the technique of endorsements peculiar to them that were still very much with us only two generations ago. At least four details we should be able to explain thus: why were there just two ceremonial gifts? In what sense did the shells serve as evidence and proof of the middlemen status of one´s immediate trade-partners? How did the distances these items have traveled add to their value? How did these gifts, unlike many shell currencies, resist inflation?

Pete Jones, The University of Manchester

Scenes as foci in the analysis of gender in film dialogue networks

Content analyses of contemporary Hollywood films consistently find a gender bias in film dialogue, with women accounting for only around 30% of speaking characters on average while also being routinely stereotyped and sexualised. The ‘Bechdel test’ is a popular device for auditing this gender bias by asking whether a film has a) two named speaking characters; who b) talk to each other; about c) something other than a man. Though the test is undoubtedly simple, it implies a relational perspective on gender bias in film, with the dialogue ties between female characters being one key site at which the biased representation of women in film can be contested. The methods of social network analysis allow us to explore this perspective through establishing films as character networks in which the gendered distribution of dialogue can be analysed. This paper outlines how films can be figured as dynamic dialogue-based character networks. I argue that the role of scenes in structuring film dialogue requires that our understanding of gendered dialogue distribution accounts for patterns of scene-sharing. Two women cannot speak to each other if they are not written into the same scene together. Moreover, it is one thing for two women to not speak to one another because they do not share a scene, and quite another for two women to share a scene but not speak to one another. The paper utilises Scott Feld’s conception of ‘foci’ in order to assess the extent to which scene-sharing structures the dialogue networks. I explore the character networks from different perspectives on aggregation: binarised networks, valued networks, as well as dynamic relational events. I also consider the multilevel cross-classification in scene-sharing as yet another level of disaggregation. I argue that film networks provide a prime example of how social ties can become focused around material social settings, and that scene-sharing is underacknowledged in the literature on gender bias in film dialogue.

Srebrenka Letina, Research Center for Educational and Network Studies, MTA, Budapest

Identifying imbalanced triads in emotional moods

Intuitively, we expect balance in our emotional experiences. When we feel happy, we tend to feel other positive emotions as well, while not feeling negative emotions in a high degree. In line with recently developed paradigm that looks at behaviors, emotions, and cognitions, not as imperfect manifestations of some unobservable latent variable, but as interacting elements of a given psychological system, our aim was to investigate this phenomenon by applying network theory to individual emotional moods. Specifically, we were interested in identifying triads that cannot be described by any dimension model because they do not satisfy the triangle inequality. For example, if moods A and B are often experienced together, and moods B and C also co-occur, then A and C are intuitively expected to be also positively correlated, while a weak tie between A and C would present forbidden strong triad, where B is a bridging mood. We hypothesized that network of moods will have more balanced – e.g. positive strong triads, and less imbalanced triads then it would be expected in the random network with the same distribution of weights.

We constructed the network of mood states based on correlation matrix of the self-reported data (Multidimensional Mood State Questionnaire) collected on the sample of 503 German-speaking adolescents and adults (aged 17 to 77) about the intensity of 69 mood states, defined as current mental state of an individual.

The network had bimodal distribution of weights, and consisted of clusters of positive and negative moods. We employed a procedure for identifying imbalanced mood triads based on the sign and strength of their links, where one edge weight is smaller than the half of other relatively strong edges in a triad, (e.g. happy, restless, and alert) and confirmed our hypothesis about the differential occurrence of certain triads in the network of moods.

Social Networks as Valuation Devices: Reputation, Ranking, Recommendations. Part 1


4 July, 17:00, Room 142

Olga Dornostup and Alena Suvorova, NRU Higher School of Economics - St. Petersburg

Community network structures of online shops' profiles on social networking site

Understanding factors led to entrepreneurial success is the attention getting area for many researchers, new firms form a ground for economic development. One of the actively explored factors related to the potential success is using the Internet tools for projects’ presentation. The aim of this study is to identify the network distinctive patterns forming the strategies of running and maintaining an online shop profile on social networking site We collected data about online shops including information about the groups followers from their profiles on We selected shops specializing in electronics and having less than 15 thousands followers, therefore, it accounts to 42 online communities in whole. We built ego graph of followers for each online shop profile and calculated the network characteristics such as mean centrality’s measures, modularity basing on fast greedy community detection algorithm and several network descriptive determinants. On the next step, we used k-means clustering algorithm to determine three clusters according to network characteristics. The largest cluster combined groups with a relatively weak network connectivity and the quite low mean share of isolated nodes. These networks are generally medium-sized and profiles are likely to have a website link, so they represented external online shop on Another cluster combined profiles with tight followers’ networks with many bridges and connected nodes along with isolated nodes. These groups are very popular and they probably use as the main source for trading. The last cluster included profiles with a small relatively well connected audience. Defining community structure is important step for further research: different types of community structure can reflect different community formation processes and product rating in smaller and more connected communities can depend more on members’ opinions than in larger communities because of stronger ties.

Elene Tabutsadze and Daria Maglevanaya, NRU Higher School of Economics - St. Petersburg

Constructing bar hierarchy in Saint-Petersburg. Institutions and reviewers evaluation structure

Our project is based on prosumerism which in recent years became an independent form of consumption. Prosumption containing both elements from production and consumption different products (Campbell 2005) also is one of the most important features of web 2.0. Web 2.0 is type of the Internet structure built on users’ activity. Hence, culture of prosuming, understood as consumer-generated system, is one of the resources of Web 2.0 (Beer et al. 2010; Shwartz 2012). The aim of our research is to define, how groups from this Internet community transpose on real bar-hopping practices and virtual evaluations of bars with following ranking. Our question refers to activity in web-application Foursquare and people, who write reviews on places which they had visited. The sample in this project consists of data from Foursquare application. We chose three main bar and restaurant streets in historical center of Saint Petersburg: Rubinstein str., Zhukovskogo str. and Dumskaya str. (Bourdieu 1989, Radayev 2016). We assume, that people, who “check-in” in certain type of locations form communities by cultural preferences (Cramer 2011; Whyte 1980). We use social network analysis to investigate how audience forms reputation of organizations. And in the same time how those visitors overlap in different bars. All chosen streets are connected with each other by users’ recommendations, however, it’s important to pay attention to the geographical characteristics and historical background. In future work this project could be used as part of recommendation system for local places of the city as an applied form of next work or it can be used as a better urban lifestyle comparative recognition of global cities in Russia.

Anastasiia Menshikova and Ilya Musabirov, NRU Higher School of Economics - St. Petersburg

Evaluation of Expertise in the Knowledge-Sharing Community: the Case of Stack Overflow

Due to the fast technological growth and professional competition on IT market, evaluation of technologies and competences is becoming more and more complicated task. However, distributed collaboration internet platforms such as GitHub and StackExchange provide us with new insights into fluent structure of professional expertise. User’s reputation is an obvious metric for professionalism assessment, but sometimes it could be not prone to voting manipulations and underestimate expertise in areas of high complexity. In this work we propose the way to estimate professional knowledge complexity with the help of network analysis on the case of Stack Overflow.

Our research aims to find the difference between extensive and intensive professional knowledge. Some users have a superficial knowledge, that is why they can answer plenty of questions in a different field (for instance, several programming languages), whereas other are able to offer a solution in narrow areas. Using linkcomm algorithm for overlapping community detection we define the skills which relate or do not relate to several communities at the same time. We discovered that those skills (or tags, in terms of Stack Overflow) which do not participate in several communities are discussed by smaller number of higher reputed users and viсe versa. Thus, metrics based on overlapping community structure could identify user expertise and evaluate knowledge complexity.

Social Media Networks


4 July, 17:00, Room 141

Tatiana Tulupyeva, St. Petersburg Institute for Informatics and Automation of the Russian Academy of Sciences (SPIIRAS); Alena Suvorova, NRU Higher School of Economics - St. Petersburg

User-generated content in social media: associations with personal traits

Information technologies are increasingly penetrating into everyday life. This leads to the fact that the Internet and social media becomes a new communication medium. User-produced content on the social networking site can present a lot of information about the personal characteristics, preferences, value orientations of the user.

Our study aimed to determine associations between type of user's posts on social network site VKontakte and his/her psychological traits. For 126 respondents we collected data about posts they published in their user profiles. For each respondent we also measured psychological defence mechanisms, value orientations, and personal traits using questionnaires (Life Style Index, The Schwartz Value Survey, Sixteen Personality Factor Questionnaire, and The Sensation Seeking Scale). All posts were classified by experts according to several features including expressing emotions (positive, negative, neutral), motivating for action, promoting, providing information about something, etc.

Our findings suggested that there was the relationship between posts features and users psychological traits. Users publishing more personal posts were more likely to be stressed and less likely to be emotionally stable. Users whose profiles included more post about different events had higher attitude towards risk. Publishing more action motivating posts was associated with higher sensation seeking. Users publishing more posts that expressed negative emotions were more likely to be anxious and intellectualization as a psychological defence mechanism. Users publishing more posts with positive emotions had higher sensation seeking score.

Social media become a great source for express diagnostics even in terms of classic methods based on questionnaire. Hence, social networks can be used by the researches as new tools for the analysis of the user’s identity, for assessing and monitoring users’ mood and emotional state.

Yadviga Sinyavskaya, Laboratory for Internet Studies, NRU Higher School of Economics - St Petersburg

Privacy attitudes and friendship in Online Social Networks

Following the discussion on the role of online social network sites in the formation of social ties, we analyze how privacy attitudes are related to friend-making behavior in social networking sites. Previous studies show that privacy attitudes may influence online behavior of users, their willingness to managing relationships or disclose personal information. It is reasonable to suppose that users tend to limit their online social circle due to privacy concerns. To the best of our knowledge, no research to date has empirically tested, how structural characteristics of ego-networks are related to the privacy attitudes of users. To address this gap, we investigate whether users with variant privacy attitudes follow different patterns of online friendship. We analyze the random sample of 368 users of largest Russian social networking site Vkontakte, who are the residents of Vologda city. Both self-reported data on privacy attitudes and the data on users’ ego-networks composition have been obtained. Different metrics of users’ ego-networks related to density, transitivity and clustering of network were calculated.

The primary look at the data reveals three groups of users divided by their privacy attitudes. It occurs that the majority of people express a little concern about several online privacy risks with strong concerned minority at hand. It turns out that privacy concerns are negatively related to density of ego-network. In addition, a series of Mann–Whitney U-test indicate difference between aforementioned groups of users: users who are highly concerned about privacy tend to form less transitive and less dense online networks.

Anna Smoliarova, Natalia Pavlushkina, and Tamara Gromova, St Petersburg State University

E-Diaspora network as spatial relations between media and their audiences

Geography and physical distance play a significant role according to the news values theory, as well as last works studying global audiences (Taneja & Webster, 2016) and global diaspora communication (Chen, Tu, & Zheng, 2017). Budarick emphasizes the difference between ethnic/migrant lenses and diaspora: the last “exist beyond the homeland or host-society and involve an array of transnational relationships between different geographical, social and cultural spaces” (Budarick, 2014, p. 144; Tsagarousianou, 2004). Myria Georgiou has developed a triangular spatial matrix for mapping diaspora media and studying transnational communication (Georgiou 2012). Global networks formed by diaspora members are seen as evidence for global public sphere (Sparks 2005, Volkmer 2007). Global networking of diaspora is well researched for ad hoc discussions in Twitter, for example, in Chinese language (Menchen-Trevino & Mao 2015, Chen, Tu, & Zheng, 2017). Network structure of Russian speaking diaspora is still underresearched. Barnett & Park (2014) have shown that the world is connected via online ties between smaller communities determined by language, geography, and historical circumstances. The studies of the web-based network include graphs visualizing relations of the most popular websites and hyperlink connections (Wu & Taneja, 2016, for Russian speaking diaspora Morgunova 2012). In this study open data about audience geography provided by is used. Statistics about countries where site's visitors located, calculating percent of visitors for each country is visualized as network by geographic locations. The findings confirm the hypothesis that Russian speaking audience in the world is not only bilaterally transnationalized but globalized. Quantitative character of the data allows to create a global geography of Russian e-diaspora and to verify other findings taken for granted in other migration studies.

Ksenia Mukhina and Alexander Visheratin, ITMO University

Events detection and monitoring using adaptive geogrids

Standard practice for event monitoring in the social network is working with a list of keywords. Using a set of keywords or hashtags, researchers are forced to predetermine what kind of events can occur in some area. However, such methods are ineffective for monitoring unexpected situations. In this scenario, the search for events should be carried out using other criteria, which are independent of the specifics of the incident.

For domain-independent event monitoring we propose two types of adaptive grids - quantitative and qualitative. Grid structure relies on a QuadTree since this representation allows improving the detection accuracy for a particular area. The main idea is to identify outliers in a spatial posting activity that are evidence of events. Initial quantifying grids are constructed according to the calendar year statistics. For each hour of working days and weekend separately, publications are placed on the city map. If the number of posts in the cell exceeds the threshold, this cell fractions until all child cells satisfy threshold conditions. Consequently, there are 576 quantifying grids for each hour of each month. A sharp rise in the number of posts inside the specific cell indicates a potential situation.

Qualitative grids are formed from a fine grid (cell size 100x100 m). Keywords, hashtags, and named entities are retrieved from publications. If the typical subjects from neighbour cells are similar, these cells unite according to QuadTree rules. Thus, the initial qualifying grid represents areas of interests in the city and the significant deviation from the typical topics in posts indicates an event.

The real-time monitoring system can adapt to current situations in case of prolonged events by using mechanism above. Experiments showed that adaptive geo-grids detect the full range of events from small concert to massive city festival.

Mixed Methods in Network Analysis. Part 1


4 July, 17:00, Room 206

Patrycja Stys, London School of Economics and Political Science

The Light in Dark Networks: Employing Mixed Methods and SNA in the Study of Governance in Fragile States

Governance in fragile, conflict-affected states challenges the Western and Westphalian ideals and ideal types of the nation-state. Fragile states experience difficulty governing territories within their borders and providing their citizens basic public welfare services like security, healthcare, and education – many of which are outsourced to non-state providers. In such states, public authority is constantly, and at times violently, negotiated between a host of actors, some unaffiliated with the formal state apparatus: from religious institutions, international agencies, and NGOs to traditional and customary leaders, non-state armed groups, and powerful business organisations. Likewise, international development initiatives and peace negotiations in such states are marred by abysmal track records of failure. Arguably, part of the problem rests in our failure to properly identify project partners and public authorities in fragile states.

This paper proposed a mixed methods approach to such identification, focusing on interviews, network data collection in the field, and the use of statistical techniques to identify with-in and –between group brokers in the provision of public welfare services in conflict-affected areas. The paper outlines (1) methodological and ethical challenges to such data collection; (2) ethnographic research design and fieldwork phases; and (3) potential uses of SNA (quantitative) and network narratives (qualitative) in analysing collected data. Negative ties and hampered service provision are also considered. The paper argues that such approaches can inform development, peace-making, and conflict resolution interventions by targeting those who are public authorities by virtual of their structural positions in social welfare service networks, as opposed to formal state representatives. The paper is based on eight months of fieldwork in rural eastern Democratic Republic of Congo (DRC) in 2016.

Daria Maltseva and Anna Shirokanova, NRU Higher School of Economics - St. Petersburg

Mixed methods in social network analysis: combining quantitative and qualitative approaches

Recently, mixed methods as a research design have become popular in the social sciences. However, there is a discussion about its novelty — the combination of different approaches is quite a usual way of doing research in sociology. Answering the critique, evangelists of mixed methods approach point out that the ‘real’ MM research should be based on integration of methods not only in data collection, but also in interpretation, and contain all the indicators of validity for qualitative and quantitative parts.

Social network analysis was institutionalized as a quantitative methodology, which uses different formal statistical and graph metrics to calculate the relationships between different actors (people, their groups or organizations) represented as networks. At the same time, there is also a tradition of qualitative approach in social network analysis. Even though originally this approach was created in 50th by social anthropologists, its development took place in late 80-90th, together with the stream of "cultural turn" in sociology and the appearance of the field called "relational sociology". Both of these perspectives have their own settled approaches to network data collection, analysis and interpretation.

Another direction of the SNA development is the integration of quantitative and qualitative approaches, which corresponds to the ‘mixed-methods strategy’ of research. The main idea of this integration is to consider the ‘dual nature of social reality’ by focusing both on network structures of relations and external contexts, on the one hand, and internal individual meanings of these relations, on the other. However, in applying the theoretical ideas of MM in practice, there arise many methodological issues on data collection, storage, analysis and interpretation.

This presentation will, first, propose a theoretical and methodological frame for further discussion on MM in SNA and discuss different characteristics that a research should have in the MM design.

Anja Znidarsic, Alenka Baggia, and Alenka Brezavscek, University of Maribor

Perception of Green Information Systems among Slovenian Managers: A Network Analysis Approach

Organizations are under constant pressure to improve their business. Beside their strategy and the legislation limitations, the important concepts that drive their decisions are sustainability and environmental issues. In the recent years, a great attention has been given to initiative of incorporating information systems into organizations that enables the sustainable development. So-called Green Information Systems (Green IS) together with environmentally-friendly green information technology play an important role also in preserving the environment.

It has been shown that the organizational culture together with personal characteristics and individual perception of managers play an important role in Green IS adoption. Therefore, it is important to investigate the perception of managers toward Green IS.

In the study, we asked managers of Slovenian enterprises what are their associations to the Green IS concept. In the first step of the analysis, a two mode network of participants times associations has been constructed. In the second step, the two mode network was transformed to a semantic one-mode network of associations. The obtained network was analysed with different social network analysis in order to investigate and classify the common views of the managers to the Green IS concept.

Ekaterina Marchenko and Denis Bulygin, NRU Higher School of Economics - St. Petersburg

Mixed methods in the analysis of team strategies: the case of Dota 2

By the term ‘mixed method’ we usually understand a combination of quantitative and qualitative methods. However, ‘mixes’ requiring reflexive methodological thinking might cross other boundaries. Social network analysis can be considered as a good example, in which mingling of different methods is appropriate for achieving better results. In our work we use a traditional task of machine learning ― association rule mining in combination with networks and clustering for analyzing team strategies in Dota 2 and their changeability related to the game patch.

Dota 2 is a noteworthy source of data because it is not only a free-to-play multiplayer online battle arena video game, but it is also one of the most influential cybersport disciplines. ESports, in its turn, became extremely popular recently, and now it is in one line with such traditional sports as soccer or American football. However, there are some significant differences between those two types of team sports. In Dota 2 more than one hundred characters to choose from in each match during the process of picks and bans, while in conventional sports strategy mostly depends on players and their positions and combinations. Besides, Dota 2 is constantly developing: its mechanics changes with every patch by introducing new heroes or altering abilities of existing ones.

In our research we reflect on the application of ML and traditional SNA methods to the complex task of team strategies analysis in Dota 2 using the case of two major tournaments (The International 2017 and DreamLeague season 8), and combining two different quantitative methods to study influence of game patch on team strategies and discuss range of teams in terms of flexibility and successful adaptability as one of predictors of team success.

Networks of International Organizations and Associations. Part 1


4 July, 17:00, Room 113

Alina Vladimirova, Institute of Oriental Studies of the Russian Academy of Sciences

Preferential Trade Agreements Network in the Asia-Pacific Region: Positional Analysis of ASEAN Member States

In 2017 the Association of Southeast Asian Nations (ASEAN) has celebrated the 50th anniversary, which led to a range of publications by scholars and journalists on this organization evolution and progress since its founding. Interesting, that assessments can differ significantly. Some authors have called ASEAN a regional leader and described its ‘remarkable strides’ in maintaining peace and accelerating economic growth across Southeast Asia. Some authors were on the opposite side and have introduced such concepts as the ‘ASEAN Drama’ in their studies on the Association development. Definitely, there is no consensus on the question if the popular notion of ‘ASEAN centrality’ represents reality or just helps to shape political discourses. Taking part in these discussions, we use network models to explore economic integration processes in the Asia-Pacific region and to test ‘ASEAN centrality’ theory. In the proposed paper we focus on preferential trade agreements network and conduct a positional analysis of ASEAN member states.

Alexander Sergunin, St. Petersburg State University

International Organizations Networks in the Arctic

This study aims to examine a network of international institutions that form a global and regional governance mechanism in the High North. The paper focuses on the role of the UN specialized agencies such as UN Development Program. UN Environment Program, International Maritime Organization, Commission on the Limits of the Continental Shelf, etc., in handling Arctic issues. Regional and subregional institutions, such as the Arctic Council, Barents-Euro-Arctic Council, Nordic institutions as well as academic/expert organizations and institutions such as International Arctic Science Committee, Intergovernmental Panel on Climate Change, etc. also will be examined.

Martin Koch, Bielefeld University

Networking in World Politics: The Role of G20

The G20 appears to be a strange animal in world politics. It is no international institution in a classical sense and no formal decision making takes place within G20. However, agreements and ideas developed within G20 have an effect on world politics. This paper deals with the following questions: (1) what are the roles and functions of G20 in world politics? (2) How and to what extent is the G20 a central node of networks in world politics and inter-organizational relations? To answer these questions the paper develops a theoretical approach based on modern systems theory that focusses on communications in world society. Using insights from the formal and informal power cycle the paper analyzes G20’s relevance in establishing an informal meeting and networking arena among world leader as means to enable formal decision-making in international organizations. Empirically the paper concentrates on documents, participatory observations of the latest G20 summit in Hamburg (Germany) and follow-up interviews in international organizations such as World Bank, IMF, WTO, OECD or the UN etc.

Anatoly Boyashov, Bielefeld University

European Union in Networks at the UN Human Rights Council

1. The UN Human Rights Council (UN HRC) is the key institution for human rights promotion in the UN system. Its aim is to promote '...universal respect for the protection of all human rights and fundamental freedoms...' [UN GA Res. 60/251, para 2].

2. To reach its aim the UN HRC functions as '...a forum... on thematic issues on all human rights...’ [UN GA Res. 60/251, para 5b]. All actors before the UN HRC prioritize certain thematic issues over the others. The European Union and its member states adopt the priorities for the UN HRC annually. The question of the paper is: how does the EU reach these priorities in practice?

3. The hypothesis of the paper is that EU establishes thematic networks to reach its priorities at the UN HRC. The EU develops two types of networks: 1) an inter-state network in the form of coalitions; 2) an inter-organizational network represented various NGOs.

4. The paper analyzes measurements of networks and visualizes two types of networks and a bipartite graph of entities pushing for the EU priorities in 2017. The inter-state network includes 28 EU member states as its nodes. The nodes are tied if the states sponsor or cosponsor a resolution at the UN HRC. The inter-organizational network includes thematic human rights issues and NGOs as its nodes. The nodes are tied if the NGOs provide a report to the OHCHR on a thematic issue. The final network includes thematic issues, states, organizations, and EU foreign policy instruments.

5. Based on the findings, the paper addresses the factors enabling or constraining the EU human rights promotion at the UN HRC.


Socio-Semantic Networks. Part 2.


5 July, 10:00, Room 140

Artem Antonyuk, Nikita Basov, and Irina Kretser, St Petersburg State University

Culture from Joint Practice: The Development of Shared Perspectives on Materiality in Creative Collectives

Symbolic interactionism shows that common physical world enables sharing of cultural meanings mediated by interaction in material contexts, especially in case of joint creative practice. At the same time, phenomenology maintains that shared meaning also emerges against the background of everyday life. Departing from these theoretical arguments, our analysis tests whether the meanings of physical reality are affected by joint practice.

The mixed-method empirical study uses the case of five European creative collectives where artists work in shared studios. We conduct a 3-mode socio-semantic-material network analysis of the interplay between joint practice of individuals and shared understanding of their everyday and creative material context. Based on the data from two waves of field research, we construct and analyze longitudinal networks of collaborations and semantic similarities between artists. To account for different types of material contexts we distinguish between two types of objects: (1) artworks, tools and materials, and (2) everyday objects, such as furniture, household objects, and food. Examining the effect of collaboration ties on semantic similarities related to shared objects enables investigating how joint practice of individuals affects their understanding of common materiality, and how this effect varies through time and across everyday and creative material contexts.

Using exponential random graph models (ERGM) for multiplex networks we find a strong tendency for collaboration ties to stimulate semantic similarity in the long term, both for creative and for everyday material contexts. Research also shows that in the short term collaborations are associated with semantic similarity with regard to creative objects, but not with regard to everyday objects. Thus, in line with the overall theoretical argument, modeling results suggest that joint material practice stimulates emergence of shared cultural meanings in both creative and everyday contexts.

Alla Loseva, NRU Higher School of Economics - St. Petersburg

Socio-semantic network of a civic association: the case of older volunteers in Saint Petersburg

My paper presents the network approach to investigating the connection between civic engagement and political participation based on interview data. There is a long-standing interest in learning what makes people switch between associational and political participation, however the most productive method so far was ethnography (c.f. Eliasoph, 1990, 1998, 2011). Narratives obtained through open-ended survey questions present an underused source of public’s political discourse yet are highly informative in terms of activists’ implicit understandings of their civic action and political participation and the relations between these spheres (Lichterman & Eliasoph, 2014, p. 817). I argue that we can study these relations by building a socio-semantic network of activists and concepts they use when describing their interaction within the organization and their personal political actions (Leifeld, 2016, 2017). First, I manually code 30 semi-structured interviews with the members of a mall volunteer association. I then reshape the corpus in the form of a network where three levels of analysis are possible, namely actors, contexts of civic or political action, and concepts related to the action. The main insight comes from the bipartite affiliation network of contexts and concepts, where the weight of a tie refers to how frequently the concept is mentioned as significant within the respective context. After projecting this network to one-mode concept-to-concept network, I extract centrality measures of concepts that connect semantic communities of civic action and political participation and conclude about the mechanisms of the connections.

Kseniia Puzyreva and Artem Antonyuk, St Petersburg State University

Patterns of community engagement in flood risk management in the south east UK: semantic analysis of local flood history narratives

The contemporary system of flood hazard management in the UK has recently experienced what policy documents label as the ‘social turn’, following an overarching shift from flood protection to flood risk management. A move towards societal flood risk management (henceforth FRM) implies the devolution of responsibilities to actors previously uninvolved in expert decision-making. These measures exemplify a broader political agenda that focuses on decentralization and promotes localism with the premise that local stakeholders are ‘those best placed to find the best solutions to local needs’. However, researchers report limited involvement and influence of citizens and communities on hazard-related decision-making. In an attempt to define conditions that influence public involvement in FRM, flood management professionals and researchers often assert flood history to be an important contextual factor influencing public engagement. This paper questions rationalist understanding of floods' influence on community engagement in FRM. Following theoretical considerations of adaptive co-management, we study how the process of knowledge formation and conceptualization of flooding events influences involvement in flood risk management. The study uses semantic network analysis of the interview data collected with representatives of community flood action groups in the village of Datchet, southeast UK in November 2017. Contrary to established understanding of the influence of flood history on community engagement in FRM, the analysis of existing conceptualizations of floods by Datchet residents shows that engagement depends not only on the event but also on its conceptualization endowed by those who experience it. The research demonstrates that different conceptualizations of seemingly the same events promote different attitudes to community engagement on the local level and lead to different action plans residents find appropriate for managing flood risks.

Qualitative Network Analysis . Part 2


5 July, 10:00, Room 213

Sophie Mützel, University of Lucerne

The emergence of a new global scientific category: a cultural analysis

Networks are composed of “culturally constituted processes of communicative interactions” (Mische) across heterogeneous actors. Empirically, this theoretical insight translates into tracing stories actors tell about themselves and others, to establish their linkages and underlying patterns of meaning making over time. This talk uses stories as data to analyze the emergence of a new scientific category in the development of breast cancer therapy, using a multi-method approach. It considers semantic networks, i.e. patterns of co-occurrence of terms over time, using a large data set on scientific discussions on breast cancer therapeutics spanning over 20 years. Using semantic network analytic techniques allows showing, first, shifts and drifts in category formation over time across the entire global field of breast cancer therapeutics. Second, an in-depth qualitative relational inquiry into the making of a new scientific category yields further insights on how this category came to be collaboratively constructed across a diversity of actors, especially in German, the UK and the US. Data selection and analyses purposefully allow for a look at the entangled, patterned developments in breast cancer therapy research from different though interrelated perspectives as it was happening. Based on a semantic network analysis and an inquiry into actors’ meaning-making processes, the talk suggests a cultural analysis of category construction, in which actors leverage culture to grapple with the ambiguities of newness.

Irina Antoschyuk, European University at Saint-Petersburg

Mixed method in scientific collaboration network research: exploring diaspora knowledge networks of Russian computer scientists in the UK

Studies of scientific collaboration networks are dominated by quantitative SNA, focusing on large scale structures for the whole scientific disciplines or specialities, countries or regions. Though these studies produce a wealth of findings on macro level trends and patterns of collaboration, they are criticized for neither taking into account the diverse character and meaning of collaborative ties nor explaining the dynamics of network building and evolution of the network. Therefore there is a growing recognition that scientific networks should be explored with a mixed methodology or integrated approach, incorporating quantitative and qualitative techniques (Lievrouw 1990; Velden, Lagoze 2013). Seeking to contribute to the discussion on how to achieve an effective combination of quantitative and qualitative methods, I share my experience of investigating diaspora knowledge networks (DKN) of Russian-speaking computer scientists (RCS) in the UK. Using various data sources, including DBLP computer science bibliography (publications and co-authorship data) and complementing it with semi-structured interviews, university websites and professional social networks (biographical information), I demonstrate that combination of methods might be fruitful on both the data collection as well as data analysis stage, forming consecutive cycles of enquiry. Thus, my interviews were collected in two portions, where the number and sample of the respondents was based on the quantitative data analysis. On the other hand, interviews analysis led to hypotheses development, which guided the statistical analysis and quantitative SNA. Finally, analysis of DKNs properties and structures as well as ego networks of individual scholars was informed and enriched by interview data. As a result, several types of ties were identified as crucially important for the network emergence and maintenance, and an explanation of the network tendency to expansion or reduction was formulated.

Joshua Eykens, University of Antwerp

A situated understanding of media ecologies and media practices in social movement studies: Network ethnography as a methodological lens

Over the past few years different disciplinary strands in the social sciences have contributed to the study of relationships between social movements and (new) media. Theoretical perspectives developed by researchers devoted to media studies however (i.e. media ecology and media practices perspectives), are urging scholars to fundamentally adjust these undertakings. In this article, we wish to join the theoretical debate that can be understood as the basis for these new perspectives. But, instead of focusing on conceptual developments, we wish to address the methodological difficulties the latter already bring. Techniques borrowed from social network analysis and ethnographic research methods have been deployed in a singularly fashion to help us better understand certain particularities of the manifold question that media scholars are currently posing. As we will see, these existing frameworks do not allow us to get a situated understanding of media ecologies as a system, with media practices taking place within them. A mixed-method approach, which has been developed by students of interorganizational settings, is being deployed and adjusted to assist us with tackling this empirical problem. The framework has been termed network ethnography and combines techniques borrowed from social network analysis with ethnographic research methods. In the following we address the theoretical duality (i.e. structure-agency) which brought us to the consideration of this mixed approach. The remainder of this contribution elaborates further on the methodological procedure. The conclusion addresses the strengths and possible pitfalls that come with such a procedure. The final section further stipulates possible ways forward for future research.

Social Networks as Valuation Devices: Reputation, Ranking, Recommendations. Part 2


5 July, 10:00, Room 142

Anastasiya Kuznetsova, NRU Higher School of Economics - St. Petersburg

Evaluating Universities via Citation and Media Networks

The evaluation of universities’ activity is traditionally a complex task due to the multidimensionality of assessment criteria. One of the promising approaches is the Webometrics Ranking of World Universities project, which creates world accepted rating based particularly on the reference of universities on the web. However, there is still need for creation of new indicators for the ranking of universities (Daraio and Bonaccorsi 2016) and this could be made through the complex implementation of webometriсs, SNA and text analysis approaches.

While web ranking methodologies like the ones used in Cybermetrics Lab (Aguillo, Ortega, and Fernández 2008) pay great attention to the analysis of university domains, we assume that the key information about universities could exist on the websites of industrial firms and mass media. Here we analyze Russian Northwestern universities according to their occurrence on the main mass media websites. We took not only simple count of references but we have also analyzed the context of their appearance on websites. Based on the topics revealed from these contexts we managed to find different clusters of universities which could be described as top-performing, regional and private universities. We found out that universities from different clusters are mentioned on different mass media websites. While top-performing universities are mentioned on mass media very often, it is a much more interesting to focus on the mentions of regional and private universities.

Vsevolod Suschevskiy and Ilya Musabirov, NRU Higher School of Economics - St. Petersburg

Network of Players Transfer in eSports. The case of Dota 2

E-sports became a major phenomenon of contemporary entertainment landscape. For sports researchers, it provides new valuable data and to some extent can serve as Castronova’s petri dish for social processes relevant both to e- and traditional sports. In this work we analyse the structure of local and regional market of player transferts. Together with a team performance estimation, this data provides us an opportunity to assess the probability of transferring between teams. Our research is based on the dataset of player transferts for the top class league and related teams, recorded from TI 16 to TI 17. We built a directed network of transferts and analyzed centralities, assortative mixing, and link formation with a help of ERGM model. The global transfer market structure shows strong geographic clusterization. One of the transfer predictors is the difference in rating, so teams which located side by side in the rating rarely have transferts.

However, transferts among the participants of “The International” major tournament are more likely inside than outside. Geographical segments not only show homophily but possess unique structure. One example of such case is Chinese e-sports ecosystem where the special structure of organizations exists with several teams under the auspices of one brand, and recruitment to high-profile teams goes through these channels.

Mixed Methods in Network Analysis. Part 2


5 July, 10:00, Room 206

Dmitry Zaytsev and Iliya Karpov, NRU Higher School of Economics - Moscow

Civil participation in Russia: from non-conventional to conventional forms (case of municipal elections in Moscow)

Mass protests become an indicator of important political transformations in Russia since “Bolotnaya movement” of 2011. It seems that mass protests emerged in 2011 and enjoyed its peaks in 2012, from the mid-2012 went down due to the repressive reaction of the government. The authors argue that the civil energy explored during the protests is not disappear but is transformed in other forms of civil participation (civil initiatives, association, social movements, online platforms and media, “monstrations”, “cultural walks”, “crowdfunding”, “occupy”, etc.). The special case is the transformation from nonconventional (strikes, demonstrations, rallies, protests) to conventional (participation in elections, advocacy campaigns, political parties, local communities, and HGOs) forms of civil participation. The authors take the case of municipal elections in Moscow of 2017 as a significant example when protests activists and politicians that are associated with 2011-2012 mass protest in Russia won elections. It was one of few successful, may be only one, example when protesters reach their goals (usually the political changes after protests was contrary to the protesters’ efforts and demands). Also this is example of transformation from nonconventional to conventional civil participation. What was the mechanism of such transformations? The authors will analyze the network of municipal deputies and their ties with civil activists and their groups related to the wave of 2011-2012 mass protests to understand to what extent the success of municipal deputies was due to the connections to the civil activists. To collect data the authors analyze social networks (Vkontakte, Facebook) with analysis of biographies and coding needed attributes. That is why methodological contribution of this paper is the use of mixed methods in network analysis of civil participation.

Daria Maltseva, NRU Higher School of Economics - St Petersburg; Stanislav Moiseev, International laboratory for Applied Network Research, NRU Higher School of Economics - Moscow

Building networks from biographical texts: different approaches to data extraction

Biographical interviews and diaries are a rich source of data for conducting network analysis. However, as a rule, these data are represented in the form of large volumes of unstructured textual information. That is why the nature of these data creates difficulties for fast and full extraction of information about the existence of connections between people, events, etc., and their characteristics.

There are three approaches to the extraction of network data – expert coding, automatic extraction, semi-automatic extraction with elements of expert coding. While expert coding allows getting the data as precise as it is needed for the research, it is time-consuming; additionally, all the members of the research group should have the same general understanding of the research aim and tasks and extract the same type of facts. With the methods of natural language processing, automatic extraction allows very fast extraction of facts and does not need a big group of researchers. However, there is a high chance not to get a full existed data as errors associated with personal names and names of organizations may occur; the process of checking such errors may take the same amount of time as expert coding. It seems like the best way to get the data from biographical interviews is to combine both approaches and add the elements of expert coding into the automatic extraction, such as pre-create a list of proper names for the search, combine different entities (for example, abbreviation and the full name of the organization), preliminary make the markup of the text.

At this presentation we will present the results of a comparison of these approaches and describe their capabilities and limitations.

Olga Silyutina, NRU Higher School of Economics - St Petersburg; Anna Shirokanova, NRU Higher School of Economics - Moscow

Mixing Social Network Analysis with Structural Topic Modelling: the case of Internet regulation coverage in the Russian media

Extracting entities from texts is one of the most crucial problems of text mining. Another problem is to find relations among those entities. When there is no prior information about the classification of the texts and links between them, building a network of those extracted entities and their connections across texts can be the basis for further analysis. One known way around is to apply the combination of topic modelling based on latent Dirichlet allocation (LDA) algorithm (Blei, 2003). However, LDA does not take into account metadata from the texts themselves such as country names. One possible solution is structural topic modeling (STM) which allows us to extract topics and take into account the information affecting the prevalence of topics or contents of the texts (Roberts et al., 2013). Getting the list of entities would produce too many categorical covariates for STM overcomplicating the model, which can be reduced with the help of clusterization conducted on the network.

In our project, we analyzed a corpus of media coverage of Internet regulation, 2009-2017, produced by Russian media. We collected additional data on the outlets’ main subjects and periodicity from their official web sites. Russian mass media cover not only local but also foreign experience of Internet regulation. In this research we made a query request for publications on the regulation and control of the Internet data from Integrum, one of the largest databases of Russian periodicals, to get their texts and metadata about their type. We extracted names of the countries mentioned together in each of the 7,400 texts. Then we projected the bipartite text-country network, and ran a cluster analysis on the country vertices, which reduced the number of covariates from 48 to 5, and then proceeded to STM using these clusters and documents’ metadata as covariates. As a result, we obtained topics with prevalence of country clusters in them.

Stanislav Moiseev, International laboratory for Applied Network Research, NRU Higher School of Economics - Moscow; Daria Maltseva, NRU Higher School of Economics - St Petersburg; Anna Shirokanova, NRU Higher School of Economics - Moscow

Networks of collaboration and interaction between Russian sociologists: from biographical interviews to network analysis

We study the structure of professional community of Soviet and Russian sociologists by means of network analysis of biographical interviews. The empirical base of the project is the data of biographical interviews, collected by Boris Doktorov during the project “International Biography Initiative”. The main purpose of our study is to reconstruct networks of interactions between the key figures of the Soviet and Russian sociology. Additional methodological purposes of the study are to develop procedures of transferring data from biographical interviews in a form suitable for network analysis.

We examine different types of formal and informal networks based on professional, educational and other types of relationships. We combine the structural analysis of data with its qualitative characteristics, such as relational contexts, modality of ties and temporality, thus trying to implement an integrative approach in network analysis.

At the conference we are going to present the model we use and the methodological issues we face during combination of the two approaches. We will also present the results of the analysis of 1st generation of Soviet sociologists.

In the long run, the study will provide the basis for the allocation of generations of Russian sociologists, and will help us to trace the history of the development of sociology in the USSR and Russia. Our project contributes to the studies devoted to the analysis of the collaboration networks between scientists and complements studies of the community of Russian sociologists and its history conducted by other authors.

Networks of International Organizations and Associations. Part 2


5 July, 10:00, Room 113

James Hollway, Graduate Institute Geneva; Christoph Stadtfeld, ETH Zürich

Multilevel network dynamics and the evolution of complex environmental governance

A key challenge in global environmental politics is how to model the dynamics of complex governance systems. These systems consist of complex patterns of ties between and among actors and the institutions they establish to govern their relationship to the environment. These ties are interdependent in three ways: socially, temporally, and across levels. Dynamic Network Actor Models (DyNAMs) offer an actor-oriented statistical network model for studying the kind of time-stamped relational data that is becoming increasingly common in political science. In this paper, we argue that DyNAMs take an actor-oriented perspective that is straightforward to interpret and make full use of available temporal information to improve the precision of inference about network dependencies. We also propose an extension that enables the investigation of network dynamics across multiple levels. This enables new questions, such as when actors choose to reinforce existing ties instead of creating new ones or are influenced by historical ties. We demonstrate the value of this model using networks drawn from a novel dataset on interstate cooperation on global environmental issues that includes comprehensive information on cooperative agreements’ start and end dates.

Sergey Sаvin and Elena Moskalchuk, St Petersburg State University

Communicative strategies of Russian NGOs

The paper contains results of a research project on communicative strategies used by Russian NGOs working with United Nations. The main objective of the study is to identify configuration of communicative canals between Russian NGOs with the consultative status of the UN ECOSOC and domestic, foreign and international structures, including state and non-state actors. Our survey of the network formed by Russian NGOs and various state- and non-state structures was conducted in summer 2016. Hyperlinks of more than 600 NGO’s websites had been analyzed with the help of soft application developed by the authors. Analysis of the data carried out using the Cytoscape program showed that communicative nodes of highest centrality are formed by, first, Russian state actors, second, by organs of international intergovernmental organizations, third by foreign actors, the both state and non-state. The result of the study is to identify the four main communicative strategies and their ranking in relation to the studied simple.

Statistical Network Modelling. Part 1


5 July, 10:00, Room 141

Michael Wältermann, Georg Wolff, and Olaf Rank, University of Freiburg

Informal CEO relationships as divers of alliance formation and persistence? An intertemporal multi-relational network approach

Research in economic sociology has long stressed that economic activities are embedded in social activities and vice versa. In the recent years, a number of multi-level network studies have demonstrated these interdependencies empirically. However, due to the cross-sectional nature of most prior work, there is still hardly any evidence concerning the dynamic interplay of economic and social relations over time. While we know that ties at both levels tend to co-occur, it remains unclear which is the chicken and which one the egg – do inter-firm alliances arise from social connections between their managers or the way around? It is also unclear to what extent ties at both levels affect the long-run persistence of ties at the respective other level – in particular, are alliances more persistent if their managers are closely connected?

In this study, we address above questions by investigating how informal advice ties among CEOs affect their firms’ likelihood of forming and maintaining alliances over time. To do so, we apply Stochastic Actor-Oriented Models (SAOM) to longitudinal data collected in a German photonics cluster. Our results indicate that firms whose CEOs reported an advice link in the past are more likely to have formed or maintained an alliance meanwhile, compared to firms whose CEOs were disconnected. This also applies if we account for the effects of firms’ geographical proximity (positive) and knowledge base similarity (non-significant). Our study provides a first step towards a dynamic approach to the analysis of inter-firm collaboration at multiple levels of agency.

Georg Wolff, Michael Wältermann, and Olaf Rank, University of Freiburg

Cross-cluster linkages as policy vehicle to prevent regions from lock-in: A network analysis of cross-regional and cross-sectoral cluster cooperation in Germany

In the last decade, the promotion of cluster organizations as a policy tool of regional development experienced a rapid increase. Cluster organizations support and offer tailor-made services to their members, which are mostly companies, supporting organizations, and research institutions located close to each other. The European Union plays a pioneering role in launching regional economy support programs through cluster organizations. While the support of cluster member organizations was for long time in the focus of the subsidies, in the recent years the support of cross-cluster partnerships has come up to the agenda. The underlying idea is to facilitate knowledge and information flows between distinct regions and sectors to prevent local economies from economic as well as cognitive lock-in effects. However, up to now, little is known about the extent and the drivers of these collaborations. For that purpose, we collected a unique dataset from 93 leading clusters operating across different industries and regions in Germany. Subsequently, we investigated how cluster organizations’ proximity in terms of knowledge bases and geographic locations affect their collaboration patterns. In doing so, we utilize Exponential Random Graph Models to estimate the impact of both forms of proximity jointly, while taking endogenous structural effects into account. Our empirical findings reveal that cluster organizations follow a tertius iungens strategy and, as a result, create a trustful environment in which innovation can flourish. The partner selection is triggered by the geographical proximity of clusters, while knowledge relatedness of co-located clusters plays a subordinate role. Nevertheless, a related knowledge base strongly supports the creation of non-local linkages, which serve as channel for accessing novel external knowledge.

Slobodan Kacanski, Roskilde University

Interplay in corporate governance network: A multilevel network analysis of board and non-executive directorship selection in Denmark

This paper investigates the current network structure of the executive and non-executive directorship selection process in Denmark. In particular, I analyse an interplay of two mutually interdependent selection processes in order to further unfold discussion on how different interests and scarce resources over preferable social actors create dynamics in network structures. To analyse the results I applied resource dependence theory. The paper applies exponential random graph models on Danish corporate governance data over the period of five years (2010-2014) to reveal the network structure and estimate network tendencies towards the selection processes. (The analysis is still in the process, so I am yet lacking of initial results).

Tomas Diviak, University of Groningen / Charles University in Prague; Jan Kornelis Dijkstra and Tom Snijders, University of Groningen

The efficiency/security trade-off: testing a theory on criminal networks

The efficiency/security trade-off hypothesis has become prominent in the field of criminal network analysis, although it has been empirically tested in only a small number of studies. This hypothesis states that criminal networks and actors involved generally manoeuver between the immediate profit from their activities (efficiency) and working slowly towards long-term goals while remaining undetected (security). It has been argued that whether the structure of the criminal network is efficient or secure depends on the goal of the particular network. That is, networks driven by financial profit (e.g., drug traffickers) are supposed to opt for efficiency, whereas networks driven by ideology (e.g., terrorists) are supposed to opt for secure network structure. In our study we focus on five network mechanisms derived from the efficiency/security trade-off, that is, density, centralization, closure, brokerage and the balance between closure and brokerage. Specifically, we identify for each mechanism tension between individual motives of actors in the criminal network and the overall structure of the network, which may even contradict the initial individual motivation. This is tested on a sample of 11 profit-driven and 8 ideology-driven criminal networks using comparison of descriptive measures based on permutation tests and exponential random graph models (ERGMs) with their subsequent within- and between-type comparison. Results show very little support for the theory – where the theory predicts differences between these two types of criminal networks, there are either none or they are the other way around. Moreover, there are greater within-group differences than between-group differences in terms of ERGM results. Findings are discussed in the light of the refinement of theory by accounting for network dynamics in response to changing environment, individual psychological predispositions, and reformulating the theory at the micro-level.

Socio-Semantic Networks. Part 3


5 July, 12:30, Room 140

Devin James Cornell and Marcelle Cohen, University of California - Santa Barbara

Discursive Fields and Socio-semantic Networks in the Colombian Right

This work introduces a novel combination of computational and quantitative tools for the measurement and analysis of discursive fields within the Colombian political party Centro Democrático as expressed on Twitter. Network analysis quantifies discursive fields of politicians as sets of actors situating themselves ideologically through the use of discourse. This uniquely frequent use of Twitter by politicians, ideologues, and official party accounts presents a rare opportunity to learn about how politicians construct discursive fields in political institutions.

Two specific discursive frameworks are examined to explore how Centro Democrático constructs opposition to the peace agreement between the government and Centro Democrático. The first builds the legitimacy of the party using populist appeals and rhetoric from the War on Terror. The second delegitimizes the peace process through normative appeals meant to simultaneously channel and challenge traditional international legal standards around conflict. Network analysis is used to identify (a) the role of ideologues in constructing party ideology, (b) the position of politicians with respect to former president Uribe, and (c) rhetorical innovations that effectively position individuals within the field over time.

Empirical analysis is performed in several steps: (a) topic modelling is performed on the combined corpus of tweets, (b) specific discursive fields are identified as collections of related topics, (c) networks are constructed where nodes are Twitter accounts and edge weights are topic appearance similarity within topic areas, and (e) statistical models are used to capture network evolution over time to show how discursive elements predict future network properties and discursive elements. This novel investigation of discursive fields in Centro Democrático give new insight into institutional construction as a process of discourse production resulting from individuals navigating their political environment.

Valentina Baiamonte, The Graduate Institute, Geneva

Lobbying beyond policy preferences in the EU climate change and energy policy-making process: an explanation of multiple submission strategy.

The European Commission regularly opens consultations to collect position papers from interest groups and receive technical expertise that policy-makers and bureaucrats from EU institutions usually lack in the development of new policies. During these consultations, interest groups, private companies and members of the civil society can submit multiple position papers to the European Commission: individually, or together through informal coalitions, or formal umbrella organisations. Within the same consultation, however, some actors may submit multiple position papers including diverging policy preferences. What is the explanation behind this multiple submission strategy? How can we identify revealed preferences across multiple submissions? This research will combine two levels: first, the relational ties that interest groups form and, second, the policy preferences expressed by each actor. This research argues that the multiple submission of position papers is a form of lobbying strategy. Policy preferences are strategically shaped to guarantee visibility and full representation within the same consultation. A methodological approach combining network analysis and automated text analysis (factorial LDA) will allow for a more nuanced view of lobbying strategies and how policy preferences are strategically shaped by interest groups.

Qualitative Network Analysis. Part 3


5 July, 12:30, Room, 213

Diana Teloian, Schneider Electric

Energy Trade Network Analysis

The study investigates the network of natural gas trades. Nodes represent the energy trading partners. Directed links identify export and import flows. The thickness of a link is proportional to the volume of the commodity traded. Thus, the dataset of the study is weighted. Each nation's natural gas trade relationship is estimated and analysed with a node in-degree and node out-degree measures. In addition, we identify which countries have the largest betweenness centrality in the network, which are the communities present in the data. Moreover, I add different attributes to the data like continents, EU/non-EU countries and others. One of the hypothesis is that small world phenomena is present in the network. In addition, I analyze the economic and environmental pollution implications of natural gas trades. Statistical tools are used to identify distribution type of the network. I use Gephi for a network visualization and representation of the study on the world map.

Johanna Schenner, University of Vienna

What Can Global Production Network Analysis Tell About Labour Exploitation?

Recently, scholarship on labour exploitation has turned towards global production network analysis (GPNA) in order to better understand this specific process. GPNA has not always been the dominant approach to do so; indeed, in the past, both global commodity chain (GCC) and value chain (GVC) theories were commonly used to explore the root causes of labour exploitation in supply chains. While GPNA does offer additional insights in investigating how non-firm actors may influence the labour law infringements in supply chains – both GCC and GVC theories were previously criticized for their exclusive emphasis on firm-to-firm relationships to understand infringements in the workplace – scholarship has remained strangely silent on the need to take into account countries’ particularities in order to understand how the process of labour exploitation may be achieved in supply chains. This paper aims at addressing this shortcoming. In a first part, the development of GPNA is retraced by paying particular attention to the similarities and differences with both GCC and GVC theories as well as explaining why scholarship has moved on to adopting the term ‘network’. By taking the example of incidents of extreme labour exploitation in UK agriculture, the second part outlines how GPNA advances the understanding of extreme labour exploitation in contrast to GCC and GVC theories. The final part of this paper argues why it is key to use GPNA in both international and national contexts.

Laura Gherardi, Università cattolica del Sacro Cuore

Studying elite’s positional mobility and ubiquity: a theoretical and methodological proposal

In this paper I apply the notion of dominant segment (Useem and McCormack 1981) and the new notion of elites’ ubiquity to the analysis the curricula of the 149 board members of the eleven Italian firms listed in the Global Fortune 500 in 2010 (Gherardi 2014): Generali, Eni, Enel, Fiat, Unicredit Intesa SanPaolo, Telecom Italia, Poste Italiane, Finmeccanica, Premafin, Mediolanum. The dominant segment, occupying top positions in at least two different institutions, not necessarily in the industrial domain, extending the notion of interlocking directorates, contributes to élites cohesion among different fields and its revolving doors. So does ubiquity: ubiquitous are those members whose delegates occupy, in their place, top institutional positions in at least two different institutions – as it is the case, for instance, for Berlusconi: while not appearing directly on any of the eleven Boards, he is represented in all the eleven boards by one or more delegate(s). I specify then how the notion of ubiquity (1) has emerged from a wider recent qualitative research I conducted on a large sample of top managers in multinational firms based in Paris, London and Milan, international artists and global academics (Gherardi 2011; 2013; 2016) (2) allows to detect ties which are not identified by classical network analysis methodologies.

Sona Nersisyan, National Academy of Sciences of the Republic of Armenia

Inter-organizational relations in the diaspora: The case of Armenian community in Tehran

Inter-organizational relations and collaborations are important for Diaspora communities. One of their main functions is self-organization of the community through accumulation and redistribution of the social capital. According to Social Network and Diaspora Studies theoretical approaches, in this paper I try to discuss Inter-organizational relations as a phenomenon that shape "community space" focusing on ideas of locality, neighborhood, etc.. Also I discuss the issue of the location of the organizations in the community using schematic and network mapping tools.My paper examines the case of Armenian community in Tehran, where are functioning the most powerful community organizations in the Armenian Diaspora.The paper is based on study which was conducted through 3 methods: standardized interview with community members, non-participant observation, in-depth interviews with representatives of community organizations and “key members” of the community.

Statistical Network Modelling. Part 2


5 July, 12:30, Room 141

Johan Koskinen and Bella Vartanyan, University of Manchester; Vincent Lorant, Université Catholique de Louvain; Galina Daraganova, The University of Melbourne; Sten-Ake Stenberg, Stockholm University

Bayesian Hierarchical Auto-logistic Node-variable Modelling for analysing Network-level Moderation of Contagion

Drawing on the framework for accommodating local dependencies used for deriving exponential random graph, Robins, Pattison, and Elliot, (2001) proposed a social influence model for modelling binary outcomes with dependence through the network structure. This model was later extended to a general auto-logistic actor attribute model (ALAAM) and likelihood-based estimation elaborated (Daraganova, 2009; Daraganova and Robins, 2013). Here we elaborate the Bayesian inference scheme for ALAAM proposed by Koskinen (2008) to multiply observed networks by imposing a hierarchical modelling structure. The aim is twofold: firstly we allow for heterogeneity across different networks; secondly we may accommodate network-level predictors of local dependencies. For social influence or contagion broadly being defined as the tendency for individuals that are relationally tied to have a higher propensity to be similar on a binary outcome than individuals that are not directly tied, it is plausible to assume that contagion may be stronger in some contexts and weaker, or absent, in others. For example, we may expect to find that peer-influence in smoking is present in some schools but not in others. We aim here to avail researchers of the tools to find school-level or network-level determinants to explain such differences. We illustrate this approach through two applications: a historical dataset consisting of a collection of about 600 school-class networks in Sweden and a dataset on the peer effects in smoking for 11K pupils in across 6 European countries.

Moammed Saqr, Uno Fors, and Jalal Nouri, Stockholm University

Social networks and performance in a medical school

There is ample research about peer relationships and how they influence students’ academic performance using traditional descriptive methods. However, little is known about how the networks of friendships in a medical school form and what factors derive the social structure. This study was done to evaluate the factors that shaped the social structure of medical students’ networks with particular emphasis on the role of academic performance and gender differences. The ties considered in this study are the long-term, face-to-face enduring relationships. The analysis compared gender differences in two parallel sections of a medical school studying the same curriculum and in the same sociocultural context. The last year medical students were surveyed, data about age, residence, grades, socioeconomic status were recorded. We used “exponential-family random graph models” (ERGMs) implemented in Statnet R package to model the networks and identify the factors that best predict the emergence of ties between students. The male network included 69 nodes and 365 edges. The best model converged at the 5th step; besides reciprocity, triangle closure, the city of residence, out-degree and in-degree popularity; the academic performance was a significant factor both the GPA and the GPA difference between a student and his alter. In the female network, (50 nodes and 176 edges), academic performance was a not significant factor, both the GPA and the difference; while reciprocity, triangle closure, the city of residence, out-degree and in-degree popularity were. The final model in male and female network showed good goodness-of-fit statistics. These results highlight the issue of homophily on performance, as a significant factor in how males build their friendship network in contrast to females. It also emphasizes the need for better inferential models that genuinely capture the network effect on performance before jumping to conclusions using traditional descriptive models that suffer the risk of endogeneity.

Darkhan Medeuov, University of Leipzig

Proximity Revisited: Testing effects of distances on friendship

Physical distance matters for social relations. A long line of research in social network analysis attest to this intuitive link, and scholarly opinions seem to agree on that likelihood of friendship decreases with distance. Recently researchers have proceeded to detail distance effect in two interrelated ways: by unravelling non-linear dependencies between distance and relations (Preciado et al. 2012, Daraganova et al. 2012); and by considering distances other than "as the crow flies" (e.g. Sailer and McCulloh 2012). In this paper, I combine insights from both lines of research to investigate distance effects on friendship between a cohort of 157 high-school students in a city in Kazakhstan. Distinguishing between linear and walking distances, I follow Preciado et al. (2012) and explore dependence between distance and friendship assuming no precise functional form within Generalized Additive Models (GAM). I approximate explored dependence with logistic regression and include the corresponding transformation of distances into Exponential Random Graph Models to compare their explanatory power and to examine if their effects retain significance in the presence of structural and individual covariates. Besides pairwise distance, I test a role of the "access" to the city understood as the inclusion into public transport network. I construct a weighted network of bus stations and measure combined M-reach centrality of bus stations within 10-min walking distance for each actor to include its pairwise absolute difference into ERG models. Results provide only marginal support for distance hypotheses: walking distance effect tends to be somewhat stronger than that of linear distance, but both effects lose significance once structural dependencies are controlled. The difference in access, however, remains significant, suggesting that in this case mutual access to the "third places" may be more salient for friendship than pairwise proximity.

Károly Takács, Hungarian Academy of Sciences; Christoph Stadtfeld and András Vörös, ETH Zurich

The Emergence and Stability of Groups in Social Networks

One of the great puzzles in social networks research is to explain the emergence of macro-level network structures from micro-level network processes. Models have been developed that successfully link micro processes to network features such as degree distributions, small-world features or segregation. The emergence of social groups, however, is harder to link to the micro level. First, because the modularity of a network is a complex structural outcome that is characterized both by cohesion within groups and lack of connectedness between them. Second, because multiple sociological micro-level network processes jointly contribute to the explanation of how individuals form social groups and agree on their boundaries. We argue that classical social network theory that is concerned with the evolution of positive relations (forces of attraction) is not sufficient to explain the emergence of groups. Only models that additionally express the co-evolution of negative relations (forces of repulsion) are able to explain the emergence and stability of groups in social networks. We illustrate this proposal by fitting stochastic actor-oriented models (SAOMs) with theoretically grounded micro-level mechanisms to empirical data of co-evolving networks of friendship and dislike among 479 secondary-school students. We then employ the estimated micro-level model as an agent-based simulation model to investigate the emergent macro-level outcomes. We find that only models that jointly consider forces of attraction and repulsion

are able to explain the emergence and stability of groups in social networks.

Networks of the Asia-Pacific Region


5 July, 12:30, Room 206

Alina Vladimirova, Institute of Oriental Studies of the Russian Academy of Sciences; Fuad Aleskerov and Margarita Golub, NRU Higher School of Economics - Moscow

Trade Powers of the Asia-Pacific Region: A Network Analysis Approach

As new Asian power centers are rising, more and more policymakers, academics, and journalists are engaged in heated discussions on who and how is able to set norms of international cooperation in the Asia-Pacific region. They are puzzled by power relations behind the observed political and economic processes because such a highly diverse set of international actors is involved. They are interested in what strategies lead to success and achievement of a higher status in the global arena. They are arguing if blocks such as existing ASEAN or a potential ‘Quad’ can accumulate enough power to balance other countries and blocks. There are many different questions about Asian powers to answer, but it is obvious that in the modern interconnected world political power audit has to include not only tangible and intangible resources assessment, but also an analysis of international relations structure.

In our paper, we demonstrate a network analysis approach to a task of finding the most influential actors and present recently developed centralities measures that allow to create corresponding indexes of national power and account for groups influence. These network metrics are based on nodes attributes and nodes interactions, both short-range and long-range ones as well as group-wise interactions, which cannot be calculated using other notions of centrality. It is also important to mention here that even though we focus on trade powers of the Asian-Pacific region, the same analysis can be conducted with economic data on other countries and can help to explore different types of political power.

Olga Petrova, Institute of Oriental Studies of the Russian Academy of Sciences

China-Indonesia Dyad in Economic Networks of the Asia-Pacific Region

There are multiple economic networks of the Asia-Pacific region which are extremely important for oriental studies scholars to explore and analyze. We try to identify patterns that constitute and influence complex international relations in the region to be able to present analytics on particular issues in areas of security, economic and politic. We conceptualize network as a set of dyads related by incidence to implement our analysis, and in the proposed paper we examine the development of economic relations between Indonesia and China in recent years. We know that Chinese investment is particularly noted for increasing the competitiveness of Indonesia's economy, however, we believe that a broader perspective on current problems and difficulties in the relations between these two countries is needed, thus, we use network approach.

Tatiana Ermolina, Institute of Oriental Studies of the Russian Academy of Sciences

Australia and ASEAN in International Trade Networks of the Asia-Pacific Region

Australia and ASEAN have not long but progressive and interesting history of relations. Starting from 1974 these countries have developed cooperation in different areas such as security, economic and humanitarian aid. The fact is that Australia wants to be involved with ASEAN since it was established in 1967, however, there are a lot of factors influencing not only relations between Australia and ASEAN but also political situation in the Asia-Pacific region in general. Therefore Australia-ASEAN cooperation is playing a significant role, but we need to study it within the broader system of international relations. We believe network analysis methods could provide an efficient framework for our study and will help to answer a range of research questions we have now. We have started to explore dynamics of Australia-ASEAN cooperation with network models of international trade and have added qualitative analysis on the development of ties between these eleven countries. As a result, we have identified important patterns which give us a better understanding of processes taking part in the Asia-Pacific region in the last 50 years.

Network Analysis of Political and Policy-Making Domains. Part 1


5 July, 12:30, Room

Nina Kolleck, Freie Universität Berlin

The role of networks in international educational and environmental politics: Uncovering influence through Social Network Analysis

Despite the relevance of education-specific negotiations under the United Nations Framework Convention on Climate Change (UNFCCC) and the influential role of the secretariat therein, research in this area is still scarce. The presentation intends to contribute to closing this research gap by exploring how the UNFCCC secretariat becomes involved in and has latent influence on the education-specific debates surrounding global climate conferences and the related information exchange on Twitter. It applies social network theory (SNT) and analysis (SNA) and derives data from Twitter to analyze the role and influence of the UNFCCC treaty secretariat within education-specific negotiations. Instead of relying on actors’ openly expressed policy preferences, their self-assessments, or their reputation for being influential, SNT and SNA infer influence from their relative position in issue-specific networks such as Twitter communication on Climate Change Education. In the last years, Twitter has increasingly been used for communication by politically influential individuals at conferences such as the climate conferences. While Climate Change Education has been one of the least prominent topics in academia for a long time, it has become a high-profile project of the UNFCCC secretariat and has steadily risen on the agenda. The presentation demonstrates evidence that with respect to Climate Change Education, the climate secretariat has increased its political influence by strategically establishing links to important actors (also beyond the formal negotiation parties), and thereby gathered support for its preferred policy options to gain a central and influential position within the education-specific communication networks in UNFCCC negotiations. The role of social networks seems to gain particular importance in policy areas like Climate Change Education which are best characterized as types of multi-level and multi-actor governance.

Ivory Mills, Northwestern University

Competing Interests: Understanding the Implications of the Emergent International ICT Governance Network

This research investigates the network of organizational actors involved in regulating ICTs throughout the international community and examines the network’s implications on competing policy interests of the vested stakeholders. Because of the unique, transnational, and convergent nature of ICT technologies and markets, international ICT governance has significant implications for consumers, service providers, device manufacturers, corporations, software developers, militaries, government agencies, and law enforcement. As such, it is important to understand international ICT governance as an example of an emergent style of governance: multistakeholder network governance, with significant and far-reaching implications as it regulates technologies that shape and transform the way we communicate and impacts the daily lives of much of the modern world. It thus attempts to fill gaps in market organization of the media and communication sector, international governance, and transnational private regulation. It utilizes social network analysis to detail and describe the governance network that shapes the ICT market, examining the emergent multi-stakeholder regulatory response that encompasses traditional international regulatory tools (treaties) and modern transnational private regulatory tools (technological standards). Additionally, it employs doctrinal and content analysis to explore the implications of this governance network on the competing policy interests, such as national security, economic development, and intellectual property protection of the various stakeholders involved – nation states, private corporations, non-governmental organizations, and even individual inventors.

Anna Mielczarek-Żejmo and Joanna Frątczak-Mueller, University of Zielona Góra

Degradation and participation. Social networks in revitalization processes

Revival in Poland is an innovative tool of social change. It manifests in new purposes (from degradation to self-organization), rules (complexity, complementarity, participation), and forms of involvement of different stakeholders. The quality of its effects depends on the strength of the innovative potential of actors holding managerial functions in government organizations, competencies of experts cooperating with them and forms of citizen involvement. The aim of the presentation is to analyze the principles of operation in the sphere of public social networks for planning social change, initiated and animated by local governments. The efficient implementation of revitalization processes encounters in Poland many barriers. Imbalance between local authorities and citizens involvement results in discontinuation of activities. or at least in focusing on selected aspects. The paradox of a statutory model of revival is expectation that politically alienated excluded people on degraded areas can be leading actors of revival. Meanwhile absence of strong non-governmental sector which can initiate and maintain changes leads the local authorities and residents to be less interested in making changes; strengthening believes that top-down methods are more efficient (principles of representative democracy); undermining believe that inhabitants are competent enough to implement changes. Diagnosed functional inefficiency of network relates to: (1) network building (civic participation in solving social problems and social control on revitalization process), (2) rationality of network flows (involvement of actors, use of public funds), (3) network effects (thriftiness and quality of public services; civic engagement).

The basis of the analyzes are revival processes carried out in 2016 in three purposely selected municipalities of Western Poland. The basis of conclusions is qualitative research (40 in-depth interviews, 24 study walks, analysis of the content of documents).

Aleksandr Sherstobitov, St Petersburg State University

Understanding Hyperdynamics, Uncertainty and Micro Level Interactions in Policy Networks: the Lessons from Quantum Physics

Network theory provides political science scholars with comprehensive analytical tools that allow to study network structure, measure wide range of network indicators and dynamics. However the policy network approach is still lacking in explanatory power as there is no universal method that enhances casual inference in understanding the networking outcomes. Moreover, the networked environment in contemporary world becomes even hyperdynamic and this is also a big challenge to network approach as it focuses basically on network structure: the configuration of ties between nodes change very quickly. The primary assumption of the proposed paper is that the ties between policy actors become less stable, more discrete and their configuration is rapidly changing. In order to study hyperdynamic networks we suggest that research focus should be shifted to micro level. We utilize the Heisenberg’s uncertainty principle from quantum physics in order to overcome the fuzziness of the modern networks nature. We argue that studying of the nodes’ ‘impulse’ may give the researcher better understanding of the policy outcomes rather than identification of the ‘frozen’ network structure and nodes’ positions at the certain period of time. The developed method focuses on the aggregate of the micro level interactions between nodes that leave ‘footprints’ such as decisions, protocols, contracts, rules of play, etc. Thus, we neglect the study of network structure in order to enhance casual inference of the policy network approach. The hypothesis and method are tested on the number of urban policy case studies in Russia.

Keynote talk

5 July, 15:30, Conference hall

John Levi Martin, University of Chicago

Elite Political Fields as Systems of Interactions: The case of the Reichstag in Weimar Germany

Sociologists resist with all their might the temptation to try to explain human action as a mere effect of structural factors, but the strength of the temptation often overpowers them. So, too, social network researchers recognize that social networks are not structures that constrain action, but are instead the cumulative result of iterated interactions. However, even where we attempt to model the logic of relationship formation and change, we tend to assume very simple actors, the sort who are easily locked in to patterns that do not reach their goal, because they are forced to follow simple rules (such as “increase your betweenness”). Skilled political actors, however, are able to break rules when it serves them, and to confound any simple analysis that confuses strategy with rule-following. The great challenge for our theories of networks is accordingly to understand political actors. We here seek to begin a close, systematic investigation of the changing logic of political interaction in the Weimar Reichstag. This is an interesting case, because shared opposition to the democratic structure led to agreement between ideologically antithetical parties. Here we examine the ways in which party members responded to one another in relatively unstructured speech. We examine whether the hostility shown to a speaker by members of another party is better predicted by their future or past voting patterns.


Statistical Network Modelling. Part 3


6 July, 10:00, Room 141

Peng Wang and Libo Liu, Swinburne University of Technology in Melbourne

The Effect of Online Social Networks on Consumer Purchase Decisions

Online consumer reviews have become increasingly prevalent on the vast majority of online shopping sites, and consumers use them either to find products that match their preferences, or to search information useful for offline purchase. The emergence of online social communities further provide platforms and channels that enhanced the dissemination of consumer opinions, which may affect consumers’ purchase decision. Traditional regression analysis have tried to predict the change in sales from a retailer’s perspective in relation to online reviews. However, how the complexity of consumer social networks structure may affect purchasing decision making process is unclear. We introduce exponential random graph models (ERGM) for social network analysis as a tool that predicts online purchasing behaviour while taking into considerations of product properties, consumer

demographics, product rating networks, as well as consumer online social networks. Testing hypothesis in relation to evidence-based and opinion-based purchase, our findings demonstrate how the multiplexed network systems may affect consumer purchasing behaviour.

Ksenia Tsyganova and Dmitri Tsyganov, St Petersburg State University

Communication vs. friendship network

The correlation between friendship network and communication network has largely been unexamined. Social relationships include complex interactions. In this paper, we only focused on measures of a communication network based on “activity” parameters. What makes an online community a friendly, safe and constructive place to join and to discuss topics of your interests? We study and compare friendship networks and communication networks in online communities. Our platform of the choice is social network site VKontacte (VK). We included additional network metrics and ERGM modeling to complement QAP tests for analyses. The main goal of this study is to compare friendship network and communication network and to find what parameters of one network can tell us about the other and vice versa. We selected several open online communities in each category: city and environment, art and music, dating and sport. From the collected data, we built several networks for each community: friendship network, comments communication network, likes communication network. In the study, we are including four network-level metrics: density, modularity, centralization, and proportion of isolates. We are also considering the following node-level metrics in modeling communication networks: degree, centrality, and betweenness. We can use this methodology to automatically classify communities by their patterns. Knowing how the attributes and patterns of friendship network affect the communication patterns, and what can be said about friendship network from knowing communication network structure, is important when the data for one or the other network is unavailable. Knowing which patterns are present in friendly and close communities and which are in hostile one, may help moderators and group organizers. What we should look for to have the productive and positive discussions in the communities and what attributes must be the warning signals that the community is degrading.

Marina Kalugina, NRU Higher School of Economics - St.Petersburg

Coalition games on networks

Game-theoretical methods became more and more often used in engineering applications, data science, economics, models of building social networks. Despite of the fact that game theory and social networks both are independent studies; the acquisition of these sciences can be very beneficial. This thesis proposes a model of social network, in which players can form a coalition and learn to get a higher payoff. Methodology we use combines game theory, social choice theory and graph theory together. The main question of this studies is to combine the knowledge of coalition formation processes and network structures, and answer the question how they affect each other. The dynamics of cooperation and the formation of coalitions is a new topic in game theory. This question is quite interesting for the Internet studies, economic, politics.

More specifically, this thesis aims to answer the question how the coalition formation influences the whole network performance and vice versa and how process of collation formation and network performance evolves in dynamics.

In our model we consider a network described by the simple two-period Romer’s model of endogenous growth with production and knowledge externalities. The sum of knowledge levels in the neighbor nodes causes an externality in the production of each node of network. The game equilibrium in the network is extensively studied. First we consider coalition formation in small network with topology like diad, triad, triangle, and star. After we expand this logic on more complex structure like a regular or a complete network with leaves (terminal vertex) attached to every node. We adopt a concept of Stochastic Shapley value for coalition formation in network with complex structure and investigate an expected payoff of different nodes in network.

Network Analysis of Political and Policy-Making Domains. Part 2


6 July, 10:00, Room 113

Halina Sapeha and Damien Contandriopoulos, University of Victoria

Structural Analysis of Health-Relevant Policy-making Information Exchange Networks in Canada

Our research project aims to understand how scientific evidence interconnects with policy-making processes. A large body of scholarship has focused on developing interventions to strengthen the influence of scientific evidence on decisions and policies. However, despite significant energy and investments, efforts to do so have proved trickier than initially anticipated. The complexity of policy-level knowledge transfer and exchange (KTE) interventions has thwarted attempts to produce strong instrumental evidence on the “how-to”. Part of the problem is rooted in the fact that much of the KTE literature focuses on discrete “interventions”. However, in practice, policy-making processes take place in complex networks where actors are interdependent and where KTE is neither linear nor discrete. Therefore, further inquiry into the composition and functioning of the channels through which information informs practices and decisions is crucial to identify best practices for fostering use of scientific evidence. Most KTE literature is based on causal attribution models, in which intervention effectiveness is conceptualized as attributable to characteristics of the strategy, users, or producers. However, if the structure of interconnections between actors is indeed a core determinant of KTE effectiveness, those attribution models are inappropriate. What becomes crucial is understanding the network structure and its functioning.

This study uses social network analysis and attempts to map and structurally analyze health-relevant policy-making networks that connect evidence production, synthesis, interpretation, and use in Canada. Research findings could strengthen the scientific understanding of how policy-level knowledge transfer and exchange functions and provide advice on how to ensure evidence plays a more prominent role in public policies.

Yuri Amirkhanian, Interdisciplinary Center for AIDS Research and Training, St Petersburg; Jeffrey Kelly, Center for AIDS Intervention Research, Medical College of Wisconsin

HIV/AIDS Prevention Interventions Impact Egocentric and Sociocentric Networks Differently: Findings of Two Large-Scale Eastern European Trials

The social network research field has made substantial contributions to benefit public health worldwide. Network sampling methods can be used to reach and recruit population members from communities that are otherwise poorly accessible. Social networks can also be used to deliver interventions, including for HIV prevention. Network data can also help understand infection disease transmission patterns. This paper discusses how a collective action framework can be applied to public health and how network interventions constitute an example of this application. Siegal, Siegal, and Bonnie (2009) have reported that, with respect to collective action, two distinct problems are identifying necessity and preserving sustainability. Two large-scale randomized intervention trials were carried out in Eastern Europe with the population of men who have sex with men. Trial 1 recruited egocentric, and Trial 2 sociocentric, networks. Risk practices were assessed prior to the intervention and then at 3- and 12-month followup points. Both trials showed significant behavioral risk reductions (Amirkhanian et al., 2005; 2015). However, Trial 1 produced only short-term positive outcomes that were not sustained, while Trial 2’s long-term outcomes were even stronger than the immediate outcomes. A plausible explanation is that Trial 1 recruited egocentric networks and trained a single sociometric network leader from each network as a change agent. Trial 2 recruited sociocentric networks, with multiple agents of change trained in each network. In addition, egocentric networks are rarely stable enough to produce sustainable effects. In contrast, sociocentric networks have more stable memberships because persons who migrate between egocentric networks still remain within the larger sociocentric network that is composed of these egocentric networks. Implications of these findings include consideration of the appropriate design of an action and costs of implementing a particular approach.

Galina Gradoselskaya, Ilia Karpov, and Tamara Scheglova, International laboratory for Applied Network Research, NRU Higher School of Economics - Moscow

Mapping of politically active groups on social networks of Russian regions (on the example of Karachay-Cherkess Republic)

In the report it will be shown of what segments the social and political activity on social networks in KChR consists, how widely it is provided. It is also necessary to define key groups and actors on social networks which create informal information space of the Republic.

Groups, persons and media, on different social networks were considered. In the analysis the authoring methodology on mapping of social networks at the federal and regional levels of the Russian Federation tested in several projects was used. Mapping of social networks was carried out by method of a grain clustering. Collection period: April-May, 2017. Collection networks: Facebook, VKontakte, Instagramm, Schoolmates, LiveJournal. The authoring method of a grain increment received 8 main clusters of political activity on social networks KChR (total number of groups – more than 2000, the groups devoted only KChR – more than 200): 1. Karachay-Cherkess - the Cluster includes the social and political groups devoted to KChR, media, and groups in support of opposition (and also the social movements supported by it like RKNK, Elbrusoid, etc.). The cluster is connected both to the all-Caucasian Islamic cluster, and to an oppositional cluster, and a cluster of Stavropol Territory. 2. Abkhazian. 3. Adyghe. 4. Kabardino-Balkarian. 5. General-Caucasian, Islamic - there are groups of the Caucasian republics: Chechnya, Ingushetia, Dagestan. 6. All-federal opposition. 7. Stavropol. 8. Pang-tyurkizm in the Caucasus - The cluster is based on activity of the Turkish information resources. Partially in Turkish, partially in Russian, partially in national languages of the Republic. Advance the ideas of combining of all Turkic peoples and territorial claims to the Russian. In the report each cluster explicitly will be analyzed by network methods, the most influential persons and social movements are shown, the content of their information network activity is analyzed.

Networks in Educational Environment. Part 1


6 July, 10:00, Room 140

Flóra Samu, Dorottya Kisfalusi, and Károly Takács, Hungarian Academy of Sciences

Well-Being and the Evolution of Positive and Negative Relations in School

Well-being in school is crucial for school adjustment, academic orientation, social competence, and problem behavior, but it has been studied less from the social network perspective than other school outcome variables. The extent to which pupils like being in school, however, is at least as much determined by peers (classmates) than by teachers and academic interests. In a longitudinal social network panel from Hungarian primary schools we demonstrate how friendship, latent (dislike, hate), and manifest (bullying, gossip) ties determine well-being in the bounded units of classrooms. Integration in cohesive friendship groups and relative informal status positions in the classroom are expected to play an important role for well-being. Informal status positions are enhanced by degree-related effects of popularity both in the positive and in the negative networks; the latter being the clear sign of social exclusion in social network terms. In addition, we also analyze how individual well-being acts upon social ties and lead to self-selection to isolation or oppositional groups. For these questions, we use a co-evolution model of well-being, positive, and negative ties in R-Siena.

Vera Titkova, Valeria Ivaniushina, and Daniel Alexandrov, NRU Higher School of Economics – St. Petersburg

Relation between social status and academic achievement in school: analysis of friendship and antipathy networks

The relation between academic success and social status in group is not identified by researchers uniquely. Academic success does not provide overall acceptance by peer in group, but high sociometric status is often positively related with academic success. Moreover student position in peer networks is related with other individual or group characteristics: gender, ethnic status, social-economic status, bullying, risk behavior and so on.

We investigate the relation between educational achievement of students and their involvement in friendship and antipathy networks in classroom. For positive ties we study this relation in different academic contexts: in class with low, medium and high level of academic motivation. Such individual characteristics as gender, ethnic status, social-economic status are used as control variables.

Our study is based on the survey of 5058 students from 98 schools of St. Petersburg (270 classroom networks). We asked students to write down names of classmates: “With whom do you socialize most of all?” and “With whom do you socialize least of all?.” We analyze dyadic ties between students with different individual characteristics. First, we use p2 models on separate positive and negative ties. Second, we produce the p2 models for positive networks separately in class with low, medium and high academic motivation.

We find that high achievement is a strong factor of popularity among peers and a protection factor from peer antipathy. Students with the same level of grades prefer to be friends (homophily); and students with different levels of achievement avoid communication with each other. Additionally we find that the class context moderates relation between academic success and popularity (friendship network) of students. In classes with low academic motivation of peers academic achievement is negatively related to popularity.

Janina Beckmann, University of Cologne

Peer Effects and Gender Differences in Career Choices of Adolescents

Gender differences in educational and occupational choices are well documented, with girls being persistently underrepresented in science, technology, engineering and mathematics (STEM) careers in many industrialized countries. The aim of this paper is to investigate how peer effects in school shape gender stereotypical career decision-making after secondary education in Germany. The paper contributes to the emerging literature of peer influence on educational choices. While several studies find that the social environment is an important predictor of gender differences in academic achievement, fewer studies have so far looked at peer effects on educational choices of studying and working in STEM related fields. In adolescence, peers become increasingly important in conveying beliefs about appropriate male or female behavior. Accordingly, if traditional gender norms prevail in one’s peer group, girls might steer away from male-dominated occupational fields. At the same time, peers adhering to more egalitarian gender norms might act as positive role models for girls’ STEM intentions. Using longitudinal social network data on German classrooms from the Children of Immigrants Longitudinal Survey in four European Countries (CILS4EU), this paper investigates how peers’ gender norms might reinforce or weaken stereotypical occupational choices for both boys and girls. Multilevel models estimate the effects of individual and classroom peers’ gender roles on the choice of STEM related fields. Since previous research has shown that not all peers are equally influential for adolescent’s behavior, the paper also explores how peer’s gender and friend closeness moderates the impact of peer influences on gendered occupational decision-making.

Marjan Cugmas, Aleš Žiberna, and Anuška Ferligoj, University of Ljubljana

The emergence of a symmetric core-cohesive blockmodel type in interactional networks in kindergarten

The presentation addresses the emergence of a global network structure in kindergarten. The global network structure is defined with a blockmodel (a blockmodel is a network where the units are clusters of equivalent units from the studied network). The presentation consists of two parts. In the first part of the presentation, it is evaluated if the global network structure in an empirical interactional networks might be the symmetric core-cohesive blockmodel type. The symmetric core-cohesive blockmodel type consist of one group of units which are linked to all the other units in the network (popular group) and several cohesive groups of units which are internally linked to each other but they are also linked to the popular group. Using the Monte Carlo simulations, the question of whether the well-known mechanisms (popularity, transitivity and assortativity) can lead towards the symmetric core-cohesive blockmodel type (from random network) is addressed in the second part of the presentation.

New Perspectives on Science and Technology Networks. Part 1


6 July, 10:00, Room 213

Alla Loseva and Daniel Alexandrov, NRU Higher School of Economics - St.Petersburg

Cohesion in scientific fields: the cases of ecology and political science

Our paper will present the results of network analysis of different scientific fields based on bibliometric data. As recent paper claimed, "The quantitative ... investigation of the existing similarities between [Social Science and the Humanities], and Life and Hard Sciences represent the forefront of scientometrics research" (Bonaccorsi et al, 2017, p. 607). Since pioneering publications by Richard Whitley (1974, 1984) on the social and cognitive organization of science, not much empirical work was done on comparing different sciences as exemplified in very recent exchange between Richard Nelson and Richard Whitley (Nelson, 2016, Whitley, 2016). Often the comparison is done on a very general level, i.e. Social Sciences vs. Physics. We suggest to go beyong the differences between sciences and focus on the differences of narrower scientific fields within and acrosss certain disciplines, entrenched in institutional forms. We assert that within established disciplines exists the plurality of practices in very different communities of practitioners with different audiences that provide their legitimacy. We use the examples of political science and ecology as institutionally established disciplines. Both are focused on complex objects and are rather diverse in terms of methodology and connection to respective broad audiences outside of academic institutions. Using bibliometric and text data from Web of Science, we compare the network parameters of citation networks of publications and authors in different research fields within these two disciplines. We compare different research fields on global-local dimension, and use both network statistics and modeling (ERGM) to evaluate the structure of networks, international diversity and the patterns of collaboration.

Anna Keuchenius, Justus Uitermark, and Petter Tornberg, University of Amsterdam

A Community Perspective on Diffusion: The Case of Granovetter's Weak Ties Hypothesis

This paper examines how ideas change while they diffuse. As a case study, we analyze the diffusion of a specific scientific idea, namely the ‘strength of weak ties’ hypothesis, introduced by Granovetter in his 1973 paper. A network is constructed of all scholars that have referenced this particular paper, with directed edges to all other researchers that are concurrently referenced with Granovetter’s 1973 paper. We find that Granovetter’s hypothesis is used by distinct communities of scholars, each with their own key narrative into which the hypothesis is fit. The diffusion within the communities follows the S-curve typical of the diffusion of innovations. The network analysis further shows that each community is clustered around one or few hubs, i.e. scientists who are frequently referenced within their community and are responsible for carrying the hypothesis into their scientific subfield. Our interpretative analysis suggests that such hubs translate a general and ambiguous idea to fit the specific vocabularies and concerns of their communities. The larger implication of this case study is that diffusion of a scientific idea is a process in which the object itself dynamically changes in concurrence with its spread, being transformed to fit the local context of communities of people. We argue that the methodology presented in this paper has potential beyond the scientific domain, particularly in the study of diffusion of other meaning based innovations, such as opinions, behavior, and ideas.

Maria Safonova and Mikhail Sokolov, NRU Higher School of Economics - St.Petersburg

Can “dark” networks be identified on the basis of their formal properties? Studying networks of academic fraud in Russia

Can we, on the basis of formal properties only, discriminate between networks involved in certain kinds of illegal or unethical activity and their benign counterparts? A rich tradition of studies on “dark networks” and clandestine organizations suggests that the necessity to carry out secret operations creates a recognizable network topology. Most studies conducted so far relied on analyzing cases of exposed criminal or terrorist organizations; as a way of evaluating statistical significance of findings on their typology, their properties were usually compared with those of simulated networks. In this paper we suggest an alternative approach: to analyze a whole set of networks involved in certain type of activities trying to find out if those of them which allow breaches of ethics could be successfully identified on the basis of their network properties. Dissertation fraud – understood as submitting plagiarized dissertation - is a widely recognized problem in Russia. We compiled a database on defenses of 58911 higher academic degree dissertations in 2006-2011. We identified plagiarized dissertations using data of investigations of DisserNet whistleblower group looking for plagiarism in publicly available dissertations. We then reconstructed the whole network of dissertation ties between degree candidates, advisors and official “opponents” (reviewers) and tested several hypotheses on network attributes of individuals involved in academic fraud. Heckit regressions with involvement in plagiarism as the dependent variable show that formal properties of networks, such as high constraint are strongly predictive of participation in fraudulent defenses even with other parameters (discipline, location) controlled. We compare these findings with other studies of “dark networks” arguing that the precise covert network typology depends on the specific characters of illegal activity and norm enforcement in the given domain.

Shaping Social Media Discourse: The Roles of People, Institutions, Algorithms, and Other Network Agents. Part 1


6 July, 10:00, Room 206

Ivan Blekanov and Svetlana Bodrunova, St Petersburg State University

Patterns of network density in ad hoc conflictual discussions: active vs. random users in six conflicts around the world

Ad hoc discussions are gaining a growing amount of attention in scholarly discourse. But earlier research has raised doubts in comparability of ad hoc discussions in social media, as they are formed by unstable, affective, and hardly predictable issue publics.

In previous research, Twitter has demonstrated the network structure of a slightly more horizontal nature than the rest of the web, thus proving the hopes of those who saw a democratizing tool in it, but our knowledge about the structure of the discussion outbursts there is still scarce. In particular, we focus on two research questions: 1) whether the ad hoc discussions on Twitter are more horizontal in terms of relations between active and non-active users; 2) whether the discussion outbursts of similar offline nature also differ from the rest of the web in a similar way and, thus, can be identified and compared.

We have chosen inter-ethnic conflicts in the USA, Germany, France, and Russia (six cases altogether, from Ferguson riots to the attack against Charlie Hebdo) to see whether similar patterns are found in the discussion structure across countries, cases, and vocabulary sets. Choosing degree distribution as the graph proxy for differentiating discussion types, we show that, for active users, there is a power law in degree distribution and its exponent values differ from the Twitter average in the same manner across cases even if the discussion density changes. These findings are true for neutral vs. affective hashtags, as well as hashtags vs. hashtag conglomerates, while it is not true for non-active users. Thus, ad hoc discussion outbursts can be identified and compared by SNA means, and the active part of the discussion seems to be more horizontal and all-involving than Twitter on average. This adds to our knowledge on comparability of ad hoc discussions online, as well as on structural differences between core and periphery in them.

Nina Zhuravlyova, Svetlana Bodrunova, and Ivan Blekanov, St Petersburg State University

A global public sphere of compassion? Spatial expansion of #JeSuisCharlie and #JeNeSuisPasCharlie and multilingual ‘bridge users’

Within the last decade, hashtag-based publics and various aspects of the discussions produced by them have created a rapidly growing field of interdisciplinary research linking public opinion and public sphere studies to social network analysis. Despite this growth, there is still scarce evidence that ‘Habermas is on Twitter’ [Bruns & Highfield, 2016], due to the affective and non-dialogue nature of expression in social networks [Papacharissi, 2015], seemingly low capacity of ad hoc discussions to create ‘opinion crossroads’, and language boundaries that prevent, i.a., cross-cultural participation of users in a given discussion and, thus, do not let the global public sphere develop. Having this in mind, we explore the spatial dimension of two affective hashtag-based publics with mutually exclusive value-loaded positions - #JeSuisCharlie and #JeNeSuisPasCharlie. We look at language distribution within the tweet collections and the expansion of the hashtagged discussion to the languages other than French. To trace the discussion outbursts, we use automated web crawling, manual coding of tweet collections, and web graph reconstruction and visual analysis. Our results suggest that, despite the differences in the volume of expression, the language structure of both hashtags was quite similar and formed echo chambers on the level of a hashtag as well as on sub-levels. Also, we see that bilingual but not multilingual users bridge the sub-level echo chambers. We argue that global compassion publics not only lift up the idea of echo chambers to a new level but also revive the concept of spiral of silence [Noelle-Neumann, 1980].

Svetlana Bodrunova and Ivan Blekanov, St Petersburg State University

Users and user groups of political left and right in conflictual Twitter discussions in Russia, the USA and Germany

Studies of political polarization in social media so far show mixed evidence whether discussions necessarily evolve into echo-chambered cocoons or provide for opinion crossroads. Recent research shows that, for political and issue-based discussions, patterns of user clusterization may differ significantly, but cross-cultural evidence for how users polarize in issue-oriented discussions is close to non-existent. Also, for detecting user polarization, most of the studies develop network proxies for users’ political alignment, and the content of tweets is rarely taken into account. We add to the scholarly discussion by detecting user polarization based on attitudes toward political actors expressed by users in Germany, the USA, and Russia, inter-ethnic conflicts being taken as cases. We develop a mixed-method approach to detecting political grouping that includes web crawling for data collection, expert coding of users, multi-dimensional scaling, word frequency vocabulary construction, and graph visualization. Our results show that the groups detected are far from conventional left/right, and that more than two streams of political talk may co-exist in the discussion. We also show that both the debate privileging either echo chambering and the one on opinion crossroads may be misleading, as the latter is found in the discussion cores, while core/periphery axis shows clear divisions in the structure of politicized talk on Twitter. We also show that influential users lie on the crossroads of the echo chambers, despite that their institutional belonging is not expected to foster their influential and/or inter-group status.

Network Analysis of Political and Policy-Making Domains. Part 3


6 July, 14:30, Room 113

Michael Wältermann, Georg Wolff, and Olaf Rank, University of Freiburg

Cross-clustering and the role of cluster funding: A multi-level relational approach to the cooperation among public and private cluster organizations

Over the past two decades, cluster policies have become firmly established across Europe as means to promote regional competitiveness. Typically, such policies involve the establishment of cluster (management) organizations to help clusters unfolding their full potential, for example by organizing regular networking events for the cluster members. In the recent years, given the multitude of cluster organizations already in place, cluster policies at EU and national levels have increasingly shifted towards the promotion of cross-clustering. Through strategic cross-cluster collaboration – it is widely assumed – cluster organizations can help their members building external linkages needed to avoid regional lock-in. Yet, in the academic literature cross-cluster relations have been widely neglected so far.

In this study, we are particularly interested in the role of cluster organizations’ funding sources in determining their cooperative behavior. While many cluster organizations rely on governmental subsidies, others are mainly funded through their member firms. Building on principal-agent and public choice theory, we argue that these distinct modes of funding not only impact the extent and form of cross-cluster collaboration (formal partnerships or informal advice exchange). They also influence the selection of collaboration partners (publicly or privately funded). Using relational data from 93 leading cluster organizations in Germany and Exponential Random Graph Models, we investigate the effects of funding on the formation of formal partnerships between cluster organizations as well as informal advice ties between cluster managers. Our empirical results reveal that privately funded cluster organizations tend to have less formal partnerships, whereas their managers more actively seek informal advice than others. Furthermore, we find funding-based homophily at both formal and informal level, suggesting a general preference for similarly funded collaboration partners.

Elena Midler and Sofia Sorokina, Southern Federal University

Regional economy and public policy: the possibilities and limitations of network analysis in the context of ensuring competitiveness Relevance

Relevance. In the conditions of growing regional imbalances and centralization of state policy, the integration of economic agents becomes the most important factor in ensuring the country's competitiveness. The forms in which integration is currently taking place are not networked. Informal mechanisms of economic activity often supplant the formal.

The purpose of the study is to analyze the institutional structure of the regional economy from the standpoint of interaction of network structures and to identify the most effective mechanisms for coordination of key stakeholders that determine the vector of state policy.

The object of the study is a sample of the regions that are part of the Southern Federal District.

Objectives of the study:

- determine the structure of the network, the nature of the assets and the resource potential through establishing the tightness of the link between the actors;

- to reveal the degree of development of social capital and the possibility of its distribution in the system of regional economy with the subsequent measurement of relations in the group of actors under the influence of various instruments of state policy;

- Establish a dominant mechanism for coordinating various actors in the context of a given state policy vector;

- to substantiate the relationship between the mechanism of coordination of relations between subjects of relations and the possibility of increasing competitiveness.

Conclusions. In societies with a high level of social trust, networks benefit significantly from the point of view of social efficiency in comparison with integration associations of the classical type. However, in Russia the level of social trust is extremely low, and the achievement of an appropriate level of public efficiency is difficult. Therefore, vertically integrated structures, based on control over property, are still preferable.

Ingo Frank, Leibniz-Institute for East and Southeast European Studies

Visual Sensemaking of Big Event Network Data to Understand Patterns of Conflict in Post-Soviet Space

The talk will investigate by means of a case study whether plausible answers to research questions like `what role do international actors play in local conflicts?´ can be achieved by analyzing big data from publicly available sources instead of historical sources (Pellon 2013). The question to be answered via network analysis is if and to what extend Russia, i.e. the Russian government, acts as a broker between the conflict parties in the Nagorno-Karabakh conflict. The analysis is focused on the context of the outbreak of violent conflict in April 2016, but also includes the history of the conflict dynamics. CAMEO coded event data from GDELT (Leetaru & Schrodt 2013) is used as data source. The CAMEO mediation event types (Schrodt 2012) are essential to analyze the influence of governmental actors to conflict afflicted societies in complex conflict dynamics (Gerner et al. 2002). The R package relevant for relational event models (REM) is used to prepare the event data for dynamic network analysis by constructing event networks for visualization (Brandes & Lerner 2008) and hypothesis testing (Brandes et al. 2009). There are some general problems with event data based on news because of bias, e.g. selection bias. In addition news are no adequate historical sources (Dulic 2011). Furthermore there are problems with data quality of automatically coded events. Therefore the data cannot be used for valid statistical explanation. Thus I will demonstrate how political event coding and exploratory temporal network analysis can at least serve as heuristic tool to find interesting correlations and generate explanatory hypotheses towards mechanistic explanation (Biermann 2011). To this end dynamic network analysis is used to create synchronoptic views of the conflict history from multiple perspectives, i.e. the perspectives of different national newswire sources. I will show diagrams built with visone’s dynamic network layout algorithms and muxViz’s multilayer map visualization.

Aleksandr Sherstobitov and Sergey Aikhel, St Petersburg State University

Networks Blockchained: Distributed Ledger Technology as the Challenge to State and Governability

The paper focuses on the approaches to studying the phenomenon of blockchain in the context of the networked policy domain. It appears that the implementation of blockchain into public policy-making may open a Pandora’s box for the political governability as it sets a number of challenges for the State that may lose its functions and authority. It reveals new effects of network interactions: short-term verifiable iterations change the characteristics of the actors' activity and promote the redistribution of responsibility (for example, through tokenization) in conditions of increasing confidence and reducing the costs of interactions. New opportunities, among which two main ones can be distinguished. First, an option for more "precise" adjustment of the processes of collective interaction and decision-making. Second, smart contracts as a new form of contractual relations, allowing the conclusion of an agreement without the participation of a guarantor (that State used to be), inevitably causes the issue of conceptualizing the nature of changes in public policy networks. In the paper we argue that ontologically blockchain is also the new political phenomenon that transforms networked public policy domain. Therefore, adequate research tools are needed to study the new network reality. In the proposed study we attempt to adjust policy network concept considering the implementation of blockchain, evaluate the challenges to governability and make the outlook for future developments.

Networks in Educational Environment. Part 2


6 July, 14:30, Room 140

Ilya Musabirov, Alina Bakhitova, and Alina Cherepanova, NRU Higher School of Economics - St Petersburg

Friendship, advice and collaboration in blended Data Science course for non-STEM students

In this report, we analyze students’ data from 3 cohorts of a two-year blended minor specialisation on Data Science in the Higher School of Economics, St.Petersburg. We analyze two types of network: survey-based friendship network and code-sharing network based on logged actions in the virtual learning environment.

The course is interdisciplinary and most of the students especially from underrepresented programmes such as History or Oriental Studies (usually 0-3 students out of 180) do not know each other beforehand. In order to be successful in the course, students are seeking access to help and advice from peers and those on the periphery of the global friendship network may have lower opportunities to get advice. This is why it is important to analyze structure of friendship networks and its dynamic. We explore how students’ position in network affect the educational outcome. Blended nature of the specialization meaning part of the learning activity is web-based that allows us to explore new forms of interaction between students. We look at how the code-sharing between students evolve before and after we randomly assigned students to the project groups in order to force the communication and extend the friendship networks. We also examine the work of the code-sharing network in the relation to existing friendship networks and academic performance.

Karen Avanesyan, University of Vienna; Serhey Kochkin, Vladimir Kirik, Larisa Tarasenko

Conspicuous Consumption and Exclusion in Networks of the Russian Student Youth

The current socio-economic recession in Russia affects social space of agents, and it is particularly relevant how this situation influences the youth. We hypothesize that behavioral expression of the changing status dispositions in networks of the student youth is described well by the concept of conspicuous consumption. We aimed to examine how conspicuous consumption contributes to the network formation, and whether it leads to social exclusion of agents who do not demonstrate status through consumption. Based on previous research, we knew that students from big cities studying economics or law have the highest propensity to conspicuous consumption. Proceeding from this, we surveyed 150 students who met these criteria and visualized their relations in a network. It allowed us asuming that there could be a positive association between conspicuous consumption and in-degree centrality and that homophily by conspicuous consumption exists. Local polynomial regression confirmed that a higher performance of conspicuous consumption leads to a greater in-degree centrality, whereas non-conspicuous consumers have a lower prestige in the network. Moreover, community detection highlighted a network clustering as agents with the lower level of conspicuous consumption establish fewer ties. Next, we employed an ERGM method. It confirmed homophily between students by gender and conspicuous consumption. Thus, odds of establishing a relationship between conspicuous consumers are by 4.25 times higher than between the other students. Surprisingly, the model did not outline a statistically significant effect of the social class on a tie formation. Despite academic performance has a statistically significant effect, it increases the odds only by 1,3 times. The last implies that even in the groups of student youth, not academic achievement but consumer showing off has a greater role in shaping of networks.

Pnina Hirsh, Ministry of Education of Israel

From Nursing to Pre-Med in the Design of Technological (vocational) High School Tracks: Neo-Liberalization and Heterogeneous Actor-Networks

In recent years, the Israeli secondary-school system has undergone various reforms. This has resulted in a gradual narrowing of the gap between academic and technological (vocational) tracks. Today, technological-track graduates can obtain a quality matriculation certificate that allows them to pursue academic studies in prestigious university departments.

This paper focuses on the health studies curriculum, its development, and its transformation from a low-status Practical Nursing track within the vocational subsystem (in the 1970s) to the prestigious Pre-Med program of the 21st century.

At the macro level, the paper analyzes the social forces that affect this transformation. Following the ANT (Actor-Network Theory) methodology, it delves into the specific details of the process, unraveling the networks that enabled the establishment of the health studies curriculum and those that led to its ‘fall from grace’ and transformation.

In the design of the Practical Nursing track we see the influence of central actors who served as policymakers in various ministries. By contrast, the Pre-Med study programs were initiated by different and separate local heterogeneous actors and networks in various social and geographical locations and spaces in Israel.

This difference reflects a shift away from a centralized educational policy that is partly the result of globalization and a neo-liberal model that allows local actors to accumulate power and influence the design of official curricula. These forces fostered the development of inter-school competition and parental choice and thus promoted efforts to attract affluent populations and high-achieving students.

The change reflects the state’s retreat from its involvement in the establishment of curricula and the increasing influence of secondary actors, a situation that might affect the equality of opportunities provided to all students.

Network Analysis of Cultural and Social Duality


6 July, 14:30, Room 142

Nadezhda Sokolova, European University at St Petersburg

Readers' literary tastes: big data for the analysis of status cultures

Cultural tastes are in the focus of studies in sociology. It was regarded as the tool of reproducing class structures (Bourdieu, 1984) and the base for drawing symbolic boundaries (Lamont and Lareau, 1988; Lamont and Molnar, 2004) as the mean of creating greater group cohesion and increasing scope of group social networks (Ericsson, 1997; Lizardo, 2006). Traditional source of data on cultural consumption are mass surveys which, however, have several important limitations. First, as a limited list of questions can be asked, most surveys use broad genre categories (e.g. subjects are asked if the like “mystery stories”), although there is general agreement that taste boundaries are often situated within, rather than between, conventional genres (Holt, 1997; Atkinson, 2011). Second, such kind of data can not reveal people' choices directly so researchers face biases in survey responses. Datasets that can be categorized as Big Data allow to avoid several previously mentioned limitations in the study of taste patterns.

The electronic dataset that was used for this research project was obtained from Saint Petersburg public library system. It contains information about readers’ literary preferences related to authors rather than any kind of conventional genre grouping and attributes such as age, gender, education and occupation. Overall, the dataset covered 2015 year has records about 1908251 books borrowed by 170312 unique readers. We (1) have studied the structure of taste patterns based on information about readers' attributes and preferences and (2) analyzed choices of various occupational groups not relying on any pre-established class or status schema identifying “culture classes” – groups of occupations demonstrating similar cultural preferences.

Mikhail Sokolov and Nadezhda Sokolova, European University at St Petersburg

A taste for centrality? A social network test of Peterson’s omnivorousness thesis using library Big Data

Testing Richard Peterson’s omnivorousness hypothesis, which states that, counter to Bourdieu, classical snobbery as a form of boundary drawing is in decline now, and elites are consuming more of both high- and low-brow art than non-elite groups, has probably been the most populated area of research in the sociology of culture in recent decades. Despite an impressive quantity of studies, an overwhelming majority of them have followed one and the same approach to proving Peterson’s proposition consisting in regressing of the number of genres consumed and/or the range of genres along the taste spectrum on various measures of status. This paper advocates an alternative approach to cultural consumption data relying on converting bi-modal networks N*K (with N standing for individuals or categories of individuals and K for objects or categories of objects) into K*K unimodal networks of objects. It is argued that major models of taste systems (such as Bourdieu’s homology, Peterson’s omnivorousness, mass culture, or individualization) have unambiguous implications for the constitution of such “culture networks” (Lizardo). We use this approach to derive and test one counterintuitive implication of Richard Peterson’s omnivorousness thesis: that it is high-brow, rather than low-brow, artistic figures that are more likely to serve as bridges across “cultural holes” (Pachucki and Breiger). We use population-level data on readership in St. Petersburg, Russia, from the city’s municipal public library system (above 1.700.000 records) to reconstruct a network of 22.000 authors, through which we demonstrate that those authors with a more educated readership are more likely to enjoy higher centrality in the taste networks measured by various measures of betweenness, closeness, and constraint. At least as far as literature is concerned, it is not “popular”, but relatively high-brow culture which is more likely to bridge cultural holes.

Nikita Basov, St Petersburg State University; Julia Brennecke, University of Liverpool; Peng Wang, Swinburne University of Technology

On the Origins of Culture: Social Networks and Emergent Meaning Structures in Small Groups

This paper investigates the social formation of culture. Drawing on interactionist reasoning, we argue culture to originate in social networks of small groups that enable engagement with group culture via usage and sharing of a group’s cultural elements. At the same time, inspired by the new structuralist movement in cultural sociology, we propose that culture arises because interacting members orchestrate relations between cultural elements and thus contribute to the emergence of group-specific meaning. We apply computer-aided text analysis and statistical network modeling to a unique mixed dataset coming from two waves of ethnographic studies conducted in five European artistic collectives. We find different social mechanisms of culture formation along two dimensions: network level (ego and dyadic) and network domain (friendship and collaboration). Although engagement with group culture is induced by social network ties across the two dimensions, only dyadic friendships enable the emergence of meaning. This study contributes to cultural sociology, social network analysis of culture, and interactionism.

Katharina Burgdorf and Hillmann Henning, University of Mannheim

Creative Revolution through Symbolic Collaboration Networks. The case of the New Hollywood Movement (1960s-1970s)

How could a small group of newcomer directors, armed with little more than a novel artistic vision (auteur theory), revolutionize the field of Hollywood filmmaking? In classical filmmaking producers and studios dominated. The new way promoted the leading role of directors. Classical narratives and visual styles were defined by temporal continuity. The new style favored non-linear narratives. Yet it is one thing to espouse a new aesthetic vision, but another to actually change the entire organization of film production. Turning artistic ideas into films requires team work to pool resources and overcome material contraints. Yet at the heart of auteur theory is the belief in the director as a unique artistic genius, the very opposite of team work. How, then, did they succeed in shifting an entire cultural field if their very identity implied a strong normative prohibition against collaboration? Drawing on longitudinal networks among 10,433 filmmakers in 6,976 film projects, we show that symbolic rather than actual collaborations within this avantgarde ensured the cohesive organization necessary to change the field of film production. Material resources were necessary to make films, and young filmmakers used the opportunity provided by Hollywood to experiment within the old studio-system. Working within commercially-driven studios threatened the movement's ideals and led to a modularized network with small clusters that ran the risk of becoming disconnected from each other, thus eroding the organizational foundation of the avantgarde. We show that a cohesive network of symbolic collaborations--shared references to their idols used by filmmakers in their own works--emerged among New Hollywood filmmakers. The dense symbolic ties fulfilled several functions at once: they were in line with the normative prohibition of actual collaboration; they signaled a shared artistic identity; and they offered a means to distinguish themselves from the established Hollywood elite.

New Perspectives on Science and Technology Networks. Part 2


6 July, 14:30, Room 213

Danica Bauer, Juan Candiani, and Victor Gilsing, University of Antwerp; Tim de Leeuw, Tilburg University

Can a network of knowledge elements adapt?

Yayavaram and Ahuja (2008) show that a nearly-decomposable knowledge base can generate more innovations than a knowledge base (henceforth KB) that is either highly decomposed (differentiated) or highly composed (integrated). A KB is conceptualized as a network of knowledge elements (nodes) and couplings (ties). Couplings originate from recombinations of knowledge elements for innovation.

In this paper, we consider the temporal effect of external knowledge sourcing on the adaption of the network structure of a focal firm’s KB. We distinguish between alliances with technological transfers, alliances without technological transfers, and inventor mobility. Alliances with technological transfers are those that enable a focal firm to acquire patents from another firm. In this process, two or more firms collaborate on a specific task and exchange explicit (in the form of patents) and tacit knowledge. In alliances without technological transfers, only tacit knowledge is exchanged. Inventor mobility is the movement of inventors between firms. Inventors embody tacit knowledge.

We argue that alliances with technological transfers can lead to a more decomposable (differentiated) KB, as knowledge elements are added to the KB but not well integrated. On the other hand, alliances without technological transfers can lead to a more non-decomposable (integrated) KB through joint learning and an ongoing exchange of tacit knowledge. Considering the movement of inventors, we argue that new inventors can establish couplings between knowledge elements more effectively, which can lead to a more non-decomposable KB (integrated).

Data was retrieved from the Compustat NA, USPTO, SDC, and Harvard Dataverse databases for 281 unique focal firms in the biopharmaceutical industry (1975-2006). The results suggest that alliances without technological transfers as well as inventor mobility indeed decompose a focal firm’s KB, while we did not find support for alliances with technological transfers.

Kamal Badar, University of Balochistan; Tatevik Poghosyan, Unu Merit, Netherlands; Julie Melville Hite, Brigham Young University Utah

The interaction of network structure, network content and absorptive capacity on firm innovation: Empirical Evidence from Armenian Board Networks

Purpose - This study, extending prior research on the effects of network resources on firm performance, examines the impact of the social capital within corporate board interlock networks on firm innovation within the context the economically transitional economy of Armenia. This research examines firm innovation based on the roles and interactions of network centrality, the debated network structures of structural holes and cohesion, network content available through board ties with partner firms, and the firm’s own absorptive capacity. Design/methodology/approach - Employing social network data of corporate board interlocks in year 2005, an innovation survey for the period 2008-2010 and the firms’ financial variables for the period 2000-2010 the hypothesized model was tested using a probit model. Findings - Findings provide empirical evidence of the influence of the interaction between network structure and network content on firm innovation. Firm absorptive capacity was significant for firm innovation when complemented by firm centrality in the board network. Implications - This study finds that while board network structure does matter, structure alone does not explain firm innovation. Rather, the combination of specific network structures and types of network content stand to provide critical value for firm innovation.

Danica Bauer, Juan Candiani, and Victor Gilsing, University of Antwerp

Internal and external network changes

In their seminal paper, Yayavaram and Ahuja (2008) show that a nearly-decomposable knowledge base can generate more innovations than a knowledge base that is either highly decomposed (decomposable/ differentiated) or highly composed (non-decomposable/ integrated). A knowledge base is conceptualized as a network of knowledge elements (nodes) and couplings (ties).

In a previous paper, we have shown that under certain circumstances knowledge networks can adapt and vary in their degree of knowledge base decomposability over time (Bauer, Candiani, Gilsing, De Leeuw, manuscript, 2017). In this paper, we study the changes of a knowledge network (variations in the degree of knowledge base decomposability) on a firm’s network position.

We argue that a firm’s network position is determined by a current state of knowledge base decomposability. If a firm’s knowledge base decomposability is high (fully decomposed), a firm’s network position is more likely to be central, as a firm may have a more diverse set of knowledge elements which can be used in collaboration agreements with (many) other firms. However, when a firm’s knowledge base decomposability decreases, this could lead to a change in a firm’s network position. For example, as a firm’s network position decreases, it cuts ties with (some) collaborating firms and may strengthen ties with others. Thus, when a knowledge base changes from a more decomposable to a more non-decomposable (more differentiated to a more integrated) knowledge base, this may lead to a decrease in centrality of a firm’s network position.

Data was retrieved from the Compustat NA, USPTO, and SDC, databases for 281 unique focal firms in the biopharmaceutical industry (1975-2006). Preliminary results confirm our arguments that changes in a firm’s internal network structure (in terms of knowledge base decomposability) go hand in hand with changes in a firm’s network position.

Shaping Social Media Discourse: The Roles of People, Institutions, Algorithms, and Other Network Agents. Part 2


6 July, 14:30, Room 206

Anna Smoliarova, St Petersburg State University

Politicians driving online discussions: are institutionalized influencers top Twitter users?

Embeddedness of politicians and political organizations in a discussion defines its level of institutionalization and creates a public arena for collaboration between publics and institutional actors. Thus, testing whether traditional hierarchies (in terms of presence of politically institutionalized actors) show up in online discussions deserves scholarly research. Moreover, it is also important to see whether more democratic societies show patterns of public involvement of politically institutionalized users that would differ from those in more authoritarian contexts.

To assess the ‘influencer’ status of politically institutionalized actors on Twitter cross-culturally, we have selected conflictual Twitter discussions in Germany, the USA, and Russia, all based on violent inter-ethnic clashes. Using vocabulary-based web crawling, we collected data on them and formed samples of top users selected by four activity metrics and five network metrics, to assess the positions of political users in the top lists and correlations of user status with their top list ranks. To this, we added qualitative assessment of presence of political users in comparative perspective. Our results show that, in all the cases, presence of political actors in online discussions is scarce; also, political actors tend to fail to link user groups or stay in the center of discussion. There is also meaningful divergence of Russia from the pattern that Germany and the USA show: while in these countries politicians gain user attention based on content, in Russia it is the status itself that matters, and political users tend to gain weight in the discussion structure despite low attention levels.

Kseniia Semykina, NRU Higher School of Economics - Moscow

LGBT-friendly and homophobes on social media in Russia: structure of friendship networks and discussions

The rights of LGBT people and the degree of freedom they should enjoy is a subject of extensive deliberation in many societies across the world. In Russia, public deliberation is hardly possible since the adoption of a law prohibiting propaganda of homosexuality to minors in 2013, deeming impossible major public discussions on the matter. In such a context, social media constitute a platform where public deliberation can be witnessed.

In this study, I attempt to understand the nature of interaction and specific features of discussions between LGBT-friendly people and homophobes by examining social networks of users who identify with one of these groups on the popular Russian social media platform VKontakte. The groups in focus are “Alliance of heterosexuals and LGBT for equality” (over 24000 participants) and “Homophobe wolf” (over 18000 followers).

The research aims at answering the following questions: 1) What are the social and geographic characteristics of the network participants (as identified in their personal information)? How are they different for the two networks? 2) What is the structure of the networks? Do the networks overlap? How active is the communication between participants of the two groups? 3) What topics are discussed in the groups? Does the discussion differ only in terms of opinions about the topics, or the topics that interest the participants are different as well? In order to analyze the structure of the groups and the nature of interaction of their participants, VkMiner software is used; and TopicMiner is used for topic modeling (both programmes developed at Higher School of Economics).

Anna Tsareva and Alexandra Radushinskaya, St Petersburg State University

Multimodal discourse analysis of strategies for managing the online audience’s attention in YouTube

The report is aimed at analyzing new forms of engaging the audience in online social media, based on the principles and strategies of the economy of attention. Various kinds of online social networks use differently the specific tools and capabilities of Internet technologies (features of production and presentation of textual and multimedia content, the network social and communication tools of “subscribing”, “following”, “sharing” and “emoticons” etc.) and involve various groups of audiences, differing in age, gender, education, income level and other characteristics. New multimodal discourses are formed on the basis of these online network resources and they combine the capabilities of modern digital technologies and specific communication strategies to attract the attention of users. One of the fastest growing online media is YouTube, the social network based on video content, that was the second most visited site in the world and had more than 1.3 billion users according to Internet statistics of 2017 year. YouTube is developing as a complex communication framework, conjoining the features of home video, mobile reporting, film channels and social and advertising media. This report presents some results of research of strategies for attracting the online social network audience’s attention, based on the data of the YouTube project "Epic Rap Battles of History". Starting at 2010 year as personal YouTube-channel of comedian Peter Shukoff, by now it contains 70 videos with more than 14 mill. subscribers, over 2,5 bil. views and has been nominated for numerous media awards. Using the multimodal discourse analysis method, the report displays two interrelated groups of analysis results: (1) the discoursive structure of the content of project’s videos as multimodal documents and (2) the set of indicators of communicative activity and interests of the explored project's audience.

Networked City: The Multiplicity of Urban Links and Nodes


6 July, 14:30, Room 141

Aleksandra Nenko and Artem Konyukhov, ITMO University

Do Places Generate Communities? Evidence from Socio-Spatial Network Analysis of Urban Data

Since M. Weber sociologists have investigated stylized way of life as a mechanism of class formation (Weber, 1922). According to P. Bourdieu social classes reproduce themselves through cultural consumption: upper and lower social classes show taste for services of elite and mass segments accordingly (Bourdieu, 1977). Another area of research has proved that public places (Jacobs, 1961) and third places (Oldenburg, 1989, 1991) are capable of generating specific milieu, where social relations are constituted and styles of life are formed. We trace these arguments through analysis of data spontaneously generated in various social networks by urban dwellers during their life in the city. We apply socio-spatial network analysis to urban data to check if spatial proximity and similarity corresponds with social proximity and similarity, as well as investigate if specific types of urban places ‘generate’ or maintain specific urban communities. Dataset is formed from social networks VKontakte, Instagram and Foursquare, which give the most vivid picture on the social connections and spatial behaviours of the users in St. Petersburg, Russia, which is the city under study. Communities are considered as clusters of user nodes connected through indirect (participation in one online public group) and direct links (friendships, mutual shares and mutual likes in social network). The person-place connection is analyzed through ‘check-ins’ retrieved from social media. Spatial networks are analyzed through geographical connections across urban venues. To trace if communities form in connection to places we track if people who share social links visit (or check-in) the same places. We target our analysis on urban places capable of community creation, i.e. ‘third places’. We account for similarity of urban venues considering popularity, style, design, average check; and for social similarity of people, considering their gender, age, interests and activity in social networks.

Alexander Visheratin, Ksenia Mukhina, and Denis Nasonov, ITMO University

Orienteering problem solving: generalized programming framework and examples for touristic trips design

Orienteering problem (OP) is a routing problem, where the aim is to generate the path through the set of nodes, which would maximize the total score and would not exceed the budget. OP was adopted for solving a number of important tasks, e.g. mobile crowdsourcing, production and tourist trip design. Since OP has received a lot of attention in last decades nowadays there is a large number of OP types - Team OP, OP with Time Windows, Generalized OP, Arc OP, etc. In this paper, we present the generalized process of solving orienteering problems and put it into practice by developing the open-source framework for solving OP tasks. Then we introduce three OP tasks related to the tourist trip design problem and demonstrate the efficiency of the developed framework through solving the described tasks using the multi-source dataset of points of interest and crowdsourced validation of the solutions.

Maria Podkorytova, St Petersburg State University

Interurban networks in post-Soviet space shaped by industrial and service companies: multiplicity and commonality

The idea of constructing interurban networks through the interaction between the service companies in accordance with the theory of central flows is quite well-known in urban hierarchy studies. However the relation between the industrial and service interurban networks is rarely studied. Meanwhile, such a comparison in the context of the former Soviet Union (FSU) space provides a new vision for studying globalisation process. Intensively globalising economics of the FSU states is mostly shaped by the Soviet industries which interlace with the global markets impacting their structures. Each FSU city used to be the core of the production and distribution chain and evolved in the specific conditions of the planned economics. Consequently, the interaction and hierarchy between the cities in FSU is strongly determined by the industrial networks.

The industrial and service networks research constitute an integral component of the regional dimension of the globalisation research. Observing the networks of international service companies we consider how global economics adopts the FSU space, while studying the networks of industrial companies we consider how the FSU space adopts global economics.

In the paper the interurban networks of the FSU cities formed by the largest industrial companies of the region (LUKoil and Gazprom) are compared to the networks constructed by the offices of the largest service companies. This unconventional approach is supposed to shade the light on the globalisation process in the FSU space and the current position of the cities within it.

Keynote talk.

6 July, 12:30, Conference hall

Robin Wagner-Pacifici, New School for Social Research

Varieties of Relational Experience in Security Strategy: Networks of Discourse

Network analysis has accomplished much but an approach to meaning remains a challenge. Drawing on previous analysis of United States National Security Strategy reports (Mohr, Wagner-Pacifici, Breiger, Bogdanov 2013; Mohr, Breiger, Wagner-Pacifici 2015; Wagner-Pacifici, Mohr and Breiger 2015; Breiger, Wagner-Pacifici, Mohr 2018), we propose to identify immanent socio-cultural templates of relational networks therein contained. Strategic policy and strategic reports issued by nation-states in the U.S., Europe, and around the globe rely on these relational templates to assess and project an international order (and dis-order) in which interactions reflect network organizing principals and roles. Deployed categories of strategic agents like friends, allies, competitors, partners, neighbors, and adversaries in the National Security Strategy reports suggest the underlying presence of multiple familiar networks (clans, neighborhoods, marketplaces, schoolyards, bureaucracies, and so forth). We seek to develop a theory and a set of methodologies to explore these hypothesized immanent networks.

Keynote talk

6 July, 17:00, Conference hall

Peter Bearman, Columbia University

The neural foundations/signatures of status and the emergence of dyadic reciprocity and transitivity in human groups

Humans are a fundamentally social species, and the social networks in which we are embedded significantly determine our physical and psychological well being, shape what is possible for us to achieve and imagine, and provide the context for social action. Given their importance and their complexity, it makes sense to think that the effectively navigating the interactions within these networks requires efficient mechanisms for processing complex multivalent social information about network members. This ability is so important that it may be among the foremost computational challenges that influenced our evolution, particularly the dramatic development of our “social brains.” This talk considers a set of findings from socializing cognitive social neuroscience that captures neural and social network data at multiple time points for interacting groups. One group involves students who volunteered to organize workers in very difficult social situations on the 50th anniversary of Freedom Summer, in the summer for respect movement. Other groups are task and leadership groups from a professional school. We believe that we can identify neural mechanisms for the reproduction of inequality in popularity in small groups. We likewise discover a truly interpersonal mechanism for the emergence of reciprocity, the building block of social solidarity. We show that we can predict from neural signatures who group members will like five months in the future. Finally, we show that we can predict from brain signatures those triples that move from intransitive to transitive over the same period.