Mapping Moral Cultures as Meaning Networks: A Computational Framework for Comparative Analysis

• In Event: 20859 - Sociology of Culture Section Roundtables
  In Refereed Roundtable (90 minute): 9. Computational Sociology & Semantic Mapping

Abstract

Amid intensifying social conflict and affective polarization, scholars increasingly agree that these divisions are rooted in moral culture: the shared understanding of what is right and wrong, or good and bad, within social groups. Yet existing research remains limited; empirical studies primarily focus on shifts in individuals’ moral value dimensions using survey measures, especially Schwartz’s value model, leaving a critical gap in understanding how these values are relationally organized and contextually interpreted. This study addresses this limitation by proposing a computational framework that reconceptualizes moral culture as a value-centered meaning network by integrating open-ended text data with Schwartz’s value model. Using a survey dataset on the 2020 U.S. presidential election (N=2,038), I combine Structural Topic Modeling (Roberts et al. 2019), the Network Comparison Test (van Borkulo et al. 2023), and Moderated Network Analysis (Haslbeck et al. 2021) to extract value-related topics and compare their relational structures between liberals and conservatives. The results indicate that while both groups exhibit comparable overall network connectivity, they differ substantially in how values structure specific political discourses. In the conservatives’ network, the Self-Transcendence/Self-Enhancement axis significantly moderates ideologically sensitive topic pairs, reinforcing in-group moral cohesion. In contrast, for liberals, these domains are organized by alternative value logics. Meanwhile, the Conservation/Openness-to-Change axis operates as a cross-ideological moderator across both groups. These findings advance research on moral culture beyond the distributional analysis of individual value scores by demonstrating how shared moral meanings can be represented and compared as networks of interconnected values and topics.

Author

• Donghyun "Henry" Kim, University of Iowa