Studying the emergent dynamics of inter-organisational net-

works can be challenging, often due to problematic empirical access.

Agent-based social simulation models can help study network phenom-

ena of interest by generating virtual networks and their dynamics. How-

ever, this approach entails an important challenge: how to implement

networking mechanisms that are a solid ontological fit with the target

context, and how to conceive of network structure as part of the mod-

elling and experimentation stages?

We use a rapid review to provide a first synthesis of papers published

in three major journals, each of them being the reference publication

for two different but connected communities: the agent based modeling

one and the social network analysis one. We point out main approaches

towards tie formation mechanisms and model choices regarding network

structural dynamics, and we highlight some major research gaps. We

conclude with a reflection on the overarching findings from the selected

articles and point out potential for future work in the simulation of emer-

gent network dynamics and required mechanisms.

Fraud in organizations is often socially embedded, emerging from interactions in which norm violations can spread and reinforce over time. Agent-based models (ABMs) provide a useful framework to study such dynamics, yet typically rely on synthetic network structures and stylized behavioral assumptions. This paper develops an empirically grounded ABM to examine how real-world organizational network structure and experimentally derived norm sensitivities jointly shape fraud diffusion. We manipulate the distribution of low norm sensitivity across the network and analyze how agents’ structural positions, local exposure, and cross-community ties shape fraud adoption. The analysis focuses on how these factors affect both the likelihood of organizational tipping and the micro-level pathways through which it unfolds. The study contributes by showing that not only the overall level of norm sensitivity and network structure shape diffusion, but also its distribution, with clustered vulnerability in specific network positions driving system-wide tipping.

The integration of biased information processing into agent-based modelling remains largely underdeveloped. To this end, we introduce a novel agent-based model grounded in pseudocontingency inference — the tendency to infer a relationship between two variables from their separate occurrences rather than their actual co-occurrences. We apply the model to the belief that healthy food is less tasty than unhealthy foods. In a first study, we validated the pseudocontingency model against experimental data manipulating the food environment. Here, we extend it to social networks and validate it against food beliefs reported in the LISS panel data from the Netherlands. The results show that the pseudocontingency model reliably reproduces observed beliefs across education groups, and that people rely more heavily on social networks than on the food environment when forming food beliefs. The pseudocontingency model thus offers a psychologically and empirically grounded approach to belief formation. The applicability of the model extends to other domains where pseudocontingency inference has been documented, including stereotype formation and consumer behaviour.

Agent-based models (ABMs) are commonly used to study the transmission of blood-borne viruses (BBVs) among people who inject drugs (PWID), yet the effects of different type of syringe-sharing partnership assumptions on the model outcomes are rarely explored.

The aim of this work is to understand the effect of two alternative assumptions of syringe-sharing partnerships on the dynamics of hepatitis C virus (HCV) among PWID. Therefore, we used an ABM of HCV transmission among PWID to compare two contrasting partnership implementations: the fixed syringe-sharing partnership in which agents select partners at the beginning of the model simulation and

maintain the same partners throughout the simulation. In dynamic proximity-based partnership, agents can select and change their partners daily based on their movement within the modelled environment and proximity to other agents. This comparison was conducted within a previously calibrated heterogeneous

model that incorporates structural heterogeneity that was implemented as differences in daily interaction frequencies among agents, and behavioural heterogeneity which was implemented as differences in syringe-sharing probabilities among the agents. All epidemiological processes, behavioural parameters, population composition, and calibration targets remain constant in the model, thus isolating the effect of the contrasting partnership types. The performance of the model was evaluated by comparing the outcomes of the model, including prevalence, incidence, cumulative uptake of treatment, cumulative cured cases,

and goodness-of-fit to the observed data measured by a mean relative error (MRE). The results indicated that the partnership assumptions meaningfully influence the model outcomes.

Specifically, dynamic proximity-based partnerships produced lower prevalence and incidence, fewer cumulative treatment initiations and cured cases, relative to fixed partnerships, despite identical sharing probabilities and interaction budgets. Differences in the fit of the model are also observed, with the dynamic proximity based configuration yielding a lower MRE. Taken together, these findings suggest that partnership implementation is not a trivial modelling assumption, but one that can shape HCV dynamics among PWID.

Rather than identifying the best partnership representation, this work highlights how partnership assumptions shape model outcomes. The findings revealed the importance of explicitly stating and testing partnership assumptions in social simulation models of infectious diseases, particularly when modelling BBV in high-risk populations.

This study examines the impact of electronic health record (EHR) sharing policies in Japan by developing an agent-based model that captures interactions among healthcare providers, patients, and the government. As population aging increases healthcare demand under persistent workforce constraints, insufficient information sharing across providers remains a major source of duplicate testing and inefficient resource use. To analyze this issue, the model incorporates providers’ decisions on electronic medical record adoption and EHR sharing, patients’ care-seeking and referral behavior, and government policies such as the introduction of a standardized EHR system and subsidies. Using CT and MRI duplication as key indicators, scenario analyses were conducted by varying the sharing rate and the timing of standardization. The results show that the effects of EHR sharing depend not only on whether sharing is introduced, but also on how quickly it diffuses and when standardization occurs. Faster diffusion substantially increases avoided duplicate tests and improves long-term cost-effectiveness, whereas delayed standardization worsens these outcomes. The findings suggest that EHR sharing should be understood not simply as a cost-reduction measure, but as a policy and institutional design issue for improving the efficient use of limited healthcare resources under asymmetric stakeholder incentives.