Recovery of resonant stochastic fluctuations in an interacting-particle system-based contagion model coupled with social mimicry: comparative analysis of the effect of event ordering in their corresponding agent based models

Invariance to the ordering of events appears to be a feature of compartmental contagion models endowed with a spatial component, which are derived from interacting particle systems (IPS). Overlaying networks on interacting particle systems appears to break that invariance by introducing path dependencies leading to different ensemble outcomes as a result of modulating the coupling between local spatial interactions and non-local information exchange. When enacted using agent-based modeling frameworks, however, the ordering of events tends to be thought of as a feature of the computational machinery used to compute model consequences rather than a choice reflecting an aspect of the physical system itself.

In this work, we show the presence and non-trivial nature of differences between various parametrizations of a coupled network-IPS contagion model, with an emphasis on two types of scheduling algorithms: a fully randomized model and a priority random model. Our model portrays an agent-based variant of the standard SIRD in which an agent’s local decisions are informed by both their local environment as well as the consensus of their spatially non-local social network. Event priorities are computed from the network centrality of each agent, leading to majority-based social mimicry, resembling the impact of social influence. Our results suggest that priority scheduling introduces resonant stochastic fluctuations modulated by social mimicry parameters, in contrast to those obtained with simple randomization. Based on the latter, we analyze the findings in the general context of path dependencies and order invariance across systems bound by local and non-local interactions. Finally we discuss how differences introduced by scheduling choices which yield significant deviations from both analytic results and phenomenological observation can provide insights on whether those discrepancies may lead to meaningful differences for scientific interpretation and decision making.