Diffusion in Colocation Contact Networks: the Impact of Nodal Spatiotemporal Dynamics

TL;DR

This paper investigates how the spatiotemporal dynamics of nodes in colocation contact networks influence spreading phenomena, using inducement-shuffling null models to analyze the effects of different correlations.

Contribution

It introduces inducement-shuffling null models to analyze the impact of spatiotemporal correlations on spreading potential in contact networks.

Findings

Node-time correlation impedes spreading.

Time-location correlation slightly promotes spreading.

Number of contacts and infection prevalence are not directly causally linked.

Abstract

Temporal contact networks are studied to understand dynamic spreading phenomena such as communicable diseases or information dissemination. To establish how spatiotemporal dynamics of nodes impact spreading potential in colocation contact networks, we propose "inducement-shuffling" null models which break one or more correlations between times, locations and nodes. By reconfiguring the time and/or location of each node's presence in the network, these models induce alternative sets of colocation events giving rise to contact networks with varying spreading potential. This enables second-order causal reasoning about how correlations in nodes' spatiotemporal preferences not only lead to a given contact network but ultimately influence the network's spreading potential. We find the correlation between nodes and times to be the greatest impediment to spreading, while the correlation between times and locations slightly catalyzes spreading. Under each of the presented null models we measure both the number of contacts and infection prevalence as a function of time, with the surprising finding that the two have no direct causality.