Conference Agenda

Causal loop diagrams (CLDs) are an increasingly popular and flexible technique that empowers us to understand the behavior of agents or factors to explain complex behavior. It is now extensively used in the social science domain since it is a tool that allows us to map attitudes and concepts and the relationships between them. These diagrams are networks; hence, their analysis benefits from the employment of tools already established in network analysis.

Several techniques can be used to construct causal loop diagrams. However, most require qualitative data collection where stakeholders are invited to identify crucial factors to represent the system and the causal relationship between them.

This study introduces a methodological contribution focused on improving the accuracy of diagram construction. The qualitative data collection and the practice of merging more than one diagram into a finalized one might be the source of several biases in the finalized diagram. For instance, a group of stakeholders might be too influential in the depiction of the system, and another group might not be represented enough. This work introduces an application of Exponential Random Graph Models (ERGMs) to systematically appraise potential biases in the diagram formalized as a network. This methodology enables researchers to explore the system’s structure and identify the underlying dynamics that led to the construction of the system under examination. The method has implications for understanding the attitudes and constructs depicted in the system.

Background: Advocacy coalitions consist of individuals who share beliefs and often act together. They have become increasingly important in community health to influence health practices. However, social network analysis has been underutilized in studying such coalitions due to the challenges of collecting whole network data across multiple settings. This paper demonstrates the use of ego-network analysis to infer coalition structures.

We apply this approach to tobacco prevention in schools. School staff are sometimes reluctant to enforce non-smoking rules, partly because they believe their colleagues do not prioritize this issue. As part of the ADHAIRE smoking prevention project, we examine homophily among teachers regarding smoking status and attitudes toward smoking prevention.

Method: We collected ego networks from 537 teachers across 20 secondary schools in a deprived Belgian province with high smoking rates. Teachers named up to five colleagues from whom they sought advice (average alters=4.7). Exponential random graph modeling (ERGM) was applied to assess homophily.

Results: We found significant homophily in attitudes toward smoking prevention (nodematch=0.525, stderr=0.008). Teachers opposed to stronger prevention measures exhibited greater homophily (nodematch=15.3) compared to those supporting more prevention (nodematch=0.33).

Conclusion: Teachers cluster into distinct coalitions based on their prevention attitudes. Opposition to prevention is particularly cohesive, weakening collective enforcement of smoking rules. Ego-network analysis offers a feasible and valuable approach to study coalition dynamics in school settings, highlighting the need to consider social divisions when implementing school-based prevention policies.

Ideological polarisation is typically understood as differences in beliefs about societal issues. We propose that polarisation can manifest not just at the level of beliefs, but at the level of how individuals perceive belief relations. A person's beliefs are embedded and structured within a broader belief system—their personal belief network. Beliefs that are not aligned with the rest of the personal network create dissonance. Previous research has explored how dissonance induces belief change (node values) but has neglected how individuals update their perceptions of the relationships between the beliefs (edge weights). We present a model of belief network formation at the individual level based on two psychological drivers: social conformity—where observing the beliefs of others influences personal belief network edges—and internal coherence—where an individual weakens/strengthens edges between dissonant/coherent beliefs. By applying this model to panel data on the political beliefs of German, Dutch, and Austrian citizens, we infer the dynamics of individuals' personal belief networks. The model predicts which beliefs are most central in individuals' networks, depending on the interplay between social and internal pressures. Our findings suggest that personal belief networks have become more interconnected and ideological in recent years, and that the average networks of political groups have diverged, reinforcing partisan divisions. Individuals polarise not only in terms of their beliefs but also more fundamentally in terms of how they structure these beliefs and what they perceive as coherent. This aspect of polarisation can widen ideological divides and undermine social cohesion.

This project seeks to map the dimensional structure underlying contemporary belief clusters in the US. Literature on American public opinion has long debated the extent to which the public is ideologically consistent. While understandings of American ideology are overwhelmingly operationalized with a liberal-to-conservative spectrum, scholars as well as recent polling data show that a plurality of Americans seem to fall outside of this dichotomy. Some recent research has then turned to belief networks and correlational analyses to understand opinion co-occurrence, without being restricted by a single-dimension ideology axis. However, this work often pre-selects highly polarized opinion areas. This can be useful to track trends in polarization, but it also artificially restricts the belief heterogeneity that can be measured. On the other hand, some recent studies use a large array of belief items, but focus on the trajectory of belief clustering (e.g., density and modularity) over time. Instead, my project focuses on the ideological contents of co-occurring beliefs. I use General Social Survey data from 2022, including all opinion questions, to inductively assess the latent dimensions organizing belief systems in the American public. I use a combination of belief network analysis – employing weighted ties to represent pair-wise absolute correlations between belief nodes – with dimensionality reduction techniques to map clusters of co-occurring beliefs and interpret the latent dimensions that organize them. The paper explores how a multidimensional ideological structure (particularly for the plurality of Independents) complicates some assumptions in works on ideological consistency and current polarization.

This study aims to explore the role of AI-generated ethical scenarios in values education with a focus on the students with different types of social networking. It attempts to divide the young students into the AI users of two types: one featured with inward-looking social networking while another with outward-looking social networking. The former is what this study described as bonding AI users while the latter as bridging AI users. Researcher would then look into whether and how these different AI users achieve certain acceptance of the positive and core values; for example, respect for others, commitment, integrity, empathy, and so on.

In order to examine the dynamic relationships between the different AI users and their formation of positive values, this study was designed to adopt qualitative research method of in-depth individual interview with the students. The research targets were Hong Kong’s students aged between 15 and 18, and interviewees comprised a total of 30 students from 5 different schools. It was generally found that by using AI-generated ethical scenarios, the bridging users could more effectively operationalize the conceptual notion of positive values as various daily-life and practical issues for their understanding. On the other hand, the bonding users could be more capable of making use of AI-generated ethical scenarios by going through the inevitable ethical reasoning and then making the decisions that were reflecting the positive and core values. Overall, this study is expected to shed some light on the innovative delivery of values education that could better address the individual needs of students with different types of social networking.