This work in progress examines the evolution of belief networks in the United States of America from the 1970s to 2022 using longitudinal data from the General Social Survey. By applying both Statistical Entailment Analysis (SEA) pioneered by Douglas White and the algebraic belief models of Martin & Wiley, we map how population-level belief networks of political, religious, and social attitudes transform over time. Following Butts and Hilgeman's approach to inferring memetic structure from cross-sectional data, we decompose observed behavioral characters into latent "microbeliefs" or "quasi-memes" that reveal underlying connections between seemingly disparate attitudes.

Preliminary findings reveal increasing polarization in belief networks on issues like abortion, with formerly nuanced combinations of microbeliefs present at the aggregate population level becoming more polarized and calcified into "all-or-nothing" positions. This contrasts with the stability observed in religious belief structures during the 1988-1998 period, suggesting differential trajectories across belief domains. Our analysis reveals complex structures that are reducible neither to distinct scales nor to models of itemwise independence, but rather form interlocking scale-like structures that evolve over time.

This approach conceptualizes beliefs as interconnected systems made up of individuals sharing complexes of microbeliefs, allowing us to work toward identifying distinct trajectories of "belief migration" and explore demographic variables that may be driving these population mental model changes in the US. By examining large, representative samples, we can investigate large-scale memetic ecologies and their evolution. Our research contributes to understanding how collective mental models evolve and how belief networks reconfigure in response to broader social changes.

This study presents findings from the project “Behavioural Change: Perspectives for the Stabilization of Sustainable Behaviours,” funded by the Università Cattolica del Sacro Cuore. Two surveys were administered to representative samples of over 2,000 Italians—first in June 2021 (when COVID had ended but its impact lingered) and again in June 2023 (when pre-pandemic routines had partially resumed).

We examined Italians’ Habits, Attitudes, and Behaviours (HAB) using Social Network Analysis to explore the correlations among approximately 50 variables (i.e. nodes) related to daily online and offline routines, sustainable practices, attitudes toward technology and pro-sociality. Our analysis detects the network structure of these HABs, focusing on layers possibly causing differences in HABs structures, i.e. gender and generations (i.e. Baby Boomers, Generation X and Generations Y and Z).

Key findings reveal that 2021 survey indicated stable HABs, with increased reliance on technology for work and leisure. The most central and strategic (measured as betweenness) HABs differ according to layers. For example, for females, sustainable habits and video chats with friends and relatives are central, while home cooking and socialising activities are more strategic in males’ networks of HABs. In 2023 the most central HABs shifted for both genders, with in-persons interactions taking a more central and strategic role, while technology-driven HABs became less central.

According to generations layers both the 2021 and 2023 surveys show different structures, with older generations changing less and younger changing more, especially in technology-related HABs, though the latter group also experienced increased anxiety and psychological distress.

Recent research has advanced cultural network analysis to map out cultural schemas held by individuals by measuring the correlations or relationality between beliefs in nationally representative surveys. However, longitudinal analysis of belief spaces is largely limited because not all beliefs were repeatedly asked over time. Since the survey questions asked multiple times are more likely to be politically charged, belief spaces constructed in this manner will likely exclude non-political and non-contentious beliefs. Our study aims to address this gap by fine-tuning large language models with the General Social Survey (GSS) from 1972 to 2021. Specifically, we analyze the latent individual belief embeddings trained during the fine-tuning process to examine the patterns of cultural belief spaces across 3,110 opinions among 68,846 individuals. Our initial analysis shows that the cultural divide between liberals and conservatives has widened, with liberals moving further to the left, whereas conservatives have maintained similar positions in the belief space from 1972 to 2021 in the GSS. We show how Americans’ cultural belief spaces are structured by socio-demographic characteristics and partisanship over time.

Recent studies on polarization have been shifting focus from elites to the mass public and from a unidimensional perspective of “polarization intensity” to a multidimensional framework encompassing both intensity and breadth. Belief Network Analysis has been applied to quantify these two dimensions. However, cross-national comparisons of opinion polarization remain limited, despite their increasing relevance in an era of deglobalization. Furthermore, the politically embedded societal context of a country may shape opinion polarization in specific ways, yet research in this area remains underexplored due to the scarcity of cross-national comparisons, which are essential for examining political influences on opinion polarization. In terms of BNA measurement, existing network indicators often reflect structural complexities, making it difficult to distill them into two unified dimensions, thereby hindering cross-national comparability.

This study utilizes the latest WVS/EVS datasets to construct belief networks for each surveyed country and employs principal component analysis to integrate network indicators into two unified polarization dimensions: the Global Concentration Index (GCI) for polarization intensity and the Universality Index (UI) for polarization breadth. We hypothesize that political environments shape polarization through top-down party competition, bottom-up civic engagement, and corresponding political-societal values. Using OLS and fixed-effects models, we find that polarization intensity is positively influenced by party competition and secular values, while the effect of party competition on polarization breadth follows a U-shaped curve. Civic engagement and self-expression values exhibit no significant impact on opinion polarization. These findings contribute to a deeper understanding of cross-national variations in opinion polarization and their institutional determinants.

This project uses a network approach to study the mental models and social learning in a small groups of stakeholders involved in a unique participatory experiment in collaborative rangeland management. Participants included traditional ranchers running an economic enterprise, conservation rangeland managers who use grazing to pursue economic and environmental goals, and government agency employees who managing public grazing programs. For two different day-long experimental sessions, these stakeholders were divided into four groups and asked to deliberate about the management of public land. Their mental models of rangeland management were measured by asking them to link their management goals to the practices that should be used to achieve the goals. This results in a bipartite network for each group, which we analysed before and after group discussion to measure social learning using temporal ERG models. We find that the most change and ‘learning’ occurred not in adding new goals and methods, but adding new relationships between the goals and methods respondents had previously mentioned. Additionally, in two of the groups, members added linkages that made their mental models significantly more similar to other group members.

Fifty years after the introduction of the European patent, the European Unified Patent Court (UPC) was established in 2023 as a unified body for the litigation of validity and infringement cases for all the participating member states in Europe. Because the UPC integrates judges from different national jurisdictions who were trained in different legal regimes and cultures, and because the UPC’s 20 divisions are geographically distributed across the 18 member states, the new court faces the challenge of offering consistent and reliable case law to businesses from around the world. We examine the social mechanisms that promote the harmonization of national patent jurisprudence within this transnational institution. After one year of operation, we conducted a network survey on over 110 technical and legal judges at the UPC to explore the extent to which they had established personal contact, read each other’s decisions and legal commentaries, and had engaged in inter-personal deliberation about general aspects of patent law before and after their appointment to the court. The findings from an explorative network analysis inform a relational, neo-structural model of transnational institutionalization that is shaped by mechanisms of both variation and convergence. Both, the appellate process as well as the formation of judicial beliefs affect harmonization. Judicial beliefs are enforced through informal deliberation networks among judges, exchanges at convergence events, scholarly commentary in publications, and citations of precedent-setting rulings. Given its geographical dispersion, the UPC will rely on the organization of temporary proximity as well as dense interpersonal deliberation networks among judges to ensure consistent jurisprudence in the future.