Publication
Contextual Network Model for Agent-Based Simulation: A Relational-Sociological Approach to Network Generation
Anna Kravets; Ye Eun Bae; Veronika Kurchyna; Philipp Flügger; Jan Ole Berndt; Ingo Timm
In: ICAART 2025, Revised Selected Papers. International Conference on Agents and Artificial Intelligence (ICAART-2025), 17th, February 23-25, Porto, Portugal, Springer, 2026.
Abstract
This paper presents the Contextual Network Model (CNM), a mechanism-driven approach to generating social networks for agent-based simulation. Grounded in formal sociology, particularly the work of Simmel, CNM formalizes tie formation through mechanisms such as reciprocity, transitivity, homophily, and contextual co-affiliation, which are embedded within layered spatial and institutional settings. In contrast to modelsthat assume fixed groups or rely on random graph structures, CNM constructs networks in which clustering and modularity emerge from patterned co-presence and interaction. Rather than proposing a universal model, CNM offers a configurable framework for generating networks suited to specific contexts where social embedding and diffusion constraints are relevant. We outline the theoretical rationale, describe the model’s generative logic, and examine its structural properties in comparison to classical network model generators. The paper concludes by identifying methodological challenges and outlining a future extension.