Conference Agenda

Network analysis aspires to be "anticategorical," yet its basic units, — relationships — are usually readily categorized entities with labels like “friendship,” “love,” or “patronage.” In this way, a nontrivial cultural typification underlies the very building blocks of most network analyses. Despite work showing that a specific “type of tie” often stands in for quite heterogeneous empirical phenomena, this typification is seldom challenged in research practice. This article expands on recent efforts to more adequately theorize ties by further developing and arguing for the concept of relationship frames — cultural models that stabilize relational expectations. I suggest that such frames are rooted in regularities in the duality of dyad and content. I then demonstrate the fruitfulness of these arguments by uncovering such regularities content of 1.2 million relationships between characters in fiction writing. Subsequently, I show how this operationalization of relationship frames allows us to conceptualize and measure tie ambiguity. The paper concludes with an exploration of which factors enhance or reduce such ambiguity.

Both social and linguistic systems often exhibit nested structures and hierarchies, which can be recovered empirically through a variety of well established hierarchical clustering techniques. However, attempts to match social and linguistic levels in socio-epistemic systems remain scarce. In this paper, we propose an information-theoretic approach to cross-correlate levels in the community and linguistic structures of a research area, which we apply to a corpus of theoretical physics publications (D=228,748 abstracts). We cluster the corpus into K=611 topics using Sentence Transformers and HDBSCAN. This linguistic partition is then organized into a dendrogram using agglomerative clustering. Additionally, we locate communities in the co-authorship network using hierarchical stochastic block modelling. Since both the linguistic and community partitions possess hierarchical structures, we may observe epistemic and social structures at arbitrary scales, from highly coarse-grained to very fine-grained social or linguistic partitions. However, at that stage, scales in these social and linguistic hierarchies are unrelated. To match communities and linguistic clusters across scales, we start by determining each scientist's ``specialty'' (the most frequent topic among their publications). These specialties can be leveraged to assess the probability that scientists belong to a certain community. Using a minimum-description length criterion, specialties are then coarse-grained adaptively by merging topics recursively (following the dendrogram structure) such as to eliminate linguistic details that add too little information about the network structure at a given scale. We validate our approach using a supervised text-classifier trained with manual annotations from an expert of the field.

Investigating the applicability of the ecological mutualism model, we consider topics and researcher communities in scientific publication networks on health as species in a mutually beneficial relationship.. Analysis of 6,430 Scopus articles on ecosystem approaches to health (Ecohealth, OneHealth, Planetary Health) reveals the co-evolution of key features of mutualistic interactions, modularity and nestedness. We compare the evolution of each approach over time, considering disruptive events and productivity. Our results show that mutualism accounts for the field's development and resilience.

Climate change has attracted new forms of activism, such as Extinction Rebellion that organizes global and local demonstrations and actions with wide-scale public attention. The focus is on how the public attention is constructed in the social networks of activist groups and their semantic networks of communication. In particular, the paper scrutinizes strategic uses of the online communications by two climate activist groups to call-for-action, spreading information, and fostering interaction around climate actions.

The analysis focuses on three types of networks; 1) social networks of participants addressing each other in online settings; 2) the semantic content of the communication between the participants, and 3) the socio-semantic networks in the activist groups’ social media posts. The paper builds upon earlier automated methods for the analysis of socio-semantic networks in social media, which use the available meta-data in social media to detect topical networks of hashtags and the words in the social media posts as well as the social networks of those authoring the posts, and those addressed in the posts.

The results show that the activist groups use their social networks to create semantic networks of support with other activist groups, and networks of opposition towards other social actors, such as individual politicians, or governmental organizations. The main strategies, calls-for-action and information spreading focus on one-way communication instead of creating a space for interaction between the different participants in the debates.

Abstract: The evolution of replicators, including DNA sequences, information embedded in linguistic formats (memes), and the output of genetic algorithms, is understood as stemming from mutation, selection, and retention. However, the overwhelming majority of mutations degrade fitness while mechanisms that inhibit deleterious mutations prevent replicators from reaching new peaks in a fitness landscape. Here we show that error correction mechanisms, in contrast to error reduction, both improve the fidelity of replicators, and accelerate progress towards a fitness peak. As a result, error correction should be strongly selected for and should be widely observable for social as well as biological phenomena. This has critical implications for the stability of information transmitted via social networks and suggests that network diffusion also represents a type of distributed network computation.

Environmental, Social, and Governance(ESG) initiative is a framework used to assess the environmental and social impact of a company and sustainability, adding the reporting of non-financial information to corporate disclosure. It helps investors make informed decisions that align with their values and long-term goals and also encourages corporations to make strategies for enhancing trust and transparency coping with related risks and even using them as opportunities for business breakthrough. This study aims to identify trends in Environmental, Social and Governance(ESG) issues in Korea and suggest future directions of academic research, corporate management strategies and support policies for ESG. This study explores social connections and meaning structures using a mixed methods of socio-semantic network analysis. The data was collected from news articles covering ESG issues and sustainability reports issued by corporations in Korea over the period from 2004 to 2024. The words in natural language were collected to identify semantic structure and interaction ties between actors were used to figure out social structure by period. Additionally, this study used Latent Dirichlet Allocation (LDA) topic modeling to extract latent themes by period and figure out the changes in the trend of ESG issues. These empirical findings suggest that the ESG issues are predominantly structured around carbon emission in environmental dimensions, human capital and safety issues in social dimensions and transparency in governance dimensions. This study Integrated both social network analysis and semantic network analysis to understand how social connections among actors and knowledge structures influence each other.

This work aims at putting to a joint empirical test the claims of such major social theories as social constructivism and symbolic interactionism. Namely, that shared meaning is established in professional and in everyday interaction, facilitated by material objects in a common physical space.

Data include collaboration and friendship ties, ethnographic descriptions of objects (artworks, materials, tools, and everyday items) filling shared spaces, and fixed individual attributes (e.g., gender and age) for three collectives of artists across three time points.

Not all objects are meaningful to every artist. Hence, any single object usually has no values on what it means to some of the artists. Furthermore, descriptions are usually short, which results in rare overlaps in semantic associations between pairs of descriptions, hindering their direct pairwise comparisons. To address this, based on descriptions of objects provided by both artists in a dyad, we used Correlated Topic Models to evaluate total dyadic similarity in meanings in dyads of artists per group/wave.

Analysing these networks is a modelling challenge due to small size, high clustering, and the nature of similarity relations derived from text and thus systematically constrained. Our pooled MRQAP models across the three groups predict total dyadic similarity from prior similarity, collaboration and friendship ties from the previous wave, and similarity in gender and age. Preliminary results suggest that social ties—particularly collaboration—predict greater similarity of meanings over time. Additionally, social closure (shared contacts) fosters similarity, while attribute similarity also plays a role, though with less consistency.