Disentangling the Role of Heterogeneity and Hyperedge Overlap in Explosive Contagion on Higher-Order Networks

Federico Malizia, Andrés Guzmán, Iacopo Iacopini, István Z. Kiss · 2025

Read Disentangling the Role of Heterogeneity and Hyperedge Overlap in Explosive Contagion on Higher-Order Networks by Federico Malizia, Andrés Guzmán, Iacopo Iacopini, István Z. Kiss (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-17800.txt.

Introduction

Federico Malizia, Andrés Guzmán, Iacopo Iacopini, and István Z. Kiss model sudden outbreak jumps on group-rich networks in a 2025 arXiv paper. Spread there runs through groups, not just pairs. Their frame tracks each group size apart, so mix and overlap stay clear. Sims back the math. The frame splits how each group size drives the whole, and shows how mix and overlap jointly set when jumps turn explosive.

Catalog record

CREATORFederico Malizia, Andrés Guzmán, Iacopo Iacopini, István Z. Kiss
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE39 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-17800

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Disentangling the Role of Heterogeneity and Hyperedge Overlap in Explosive Contagion on Higher-Order Networks Federico Malizia, Andrés Guzmán, Iacopo Iacopini, István Z. Kiss Abstract We introduce group-based compartmental modeling (GBCM), a mean-field framework for irreversible contagion in higher-order networks that captures structural heterogeneity and correlations across group sizes. Validated through numerical simulations, GBCM analytically disentangles the role of each interaction order to the global epidemic dynamics, revealing how heterogeneity and inter-order correlations jointly shape the onset of outbreaks and the emergence of explosive dynamics. Crucially, we show that inter-order correlations drive the system along distinct pathways to explosive contagion—emerging universally across both irreversible and reversible spreading processes. The propagation of contagions and behaviors in complex systems often involves repeated or simultaneous stimuli that individuals receive from their social contacts [ 1 , 2 ] . One way to mechanistically encode these non-linear effects into contagion models is to consider interactions that go beyond simple pairwise connections [ 3 , 4 , 5 ] . Recent studies have highlighted the critical role of these higher-order interactions—group involving three or more individuals—in shaping the dynamics of spreading processes [ 6 , 7 , 8 ] , but also in synchronization [ 9 , 10 , 11 ] , and game theory [ 12 , 13 , 14 , 15 ] . In fact, higher-order mechanisms [ 16 ] on complex networks give rise to a variety of new phenomena [ 17 , 18 ] , such as explosive transitions [ 19 ] , vanishing size of critical mass [ 20 ] , multi-stability [ 8 , 21 , 22 ] and chaos [ 23 ] . Crucially, it has been shown that the way these group interactions are distributed across the system plays a central role in determining its behavior [ 24 , 25 , 26 , 27 , 28 ] . From a macroscopic point of view, the presence of hubs in higher-order structures significantly impacts the onset and evolution of spreading processes [ 29 , 30 ] . More recently, the role of the microscopic arrangement of groups has been studied via two key structural concepts: intra-order correlations, describing dependencies within interactions of the same order [ 31 ] , and inter-order correlations for the interplay across orders [ 32 ] . While intra-order correlations depend on the microscopic arrangement of groups of the same size, inter-order correlations [ 33 ] depend on the full hierarchy of groups, and can profoundly influence collective dynamics. For instance, higher-order networks with uncorrelated sets of hyperedges behave fundamentally differently from those structured as simplicial complexes—where the downward closure requirement maximizes correlations across orders [ 34 , 35 , 32 ] . Although significant progress has been made in characterizing the impact of structural network properties on the contagion dynamics that unfolds over it [ 29 , 24 , 36 , 30 , 8 , 28 , 32 , 31 ] , capturing at the same time inter- and intra-order correlations, hyperdegree distributions, and state dependencies across different orders of interactions remains a challenge. These features introduce substantial complexity, and existing frameworks often fail to balance analytical tractability with the need to account for both structural and dynamical heterogeneity [ 37 , 32 ] . In this paper, we develop a group-based mean-field framework to study the dynamics of the Susceptible-Infected-Recovered (SIR) model on higher-order networks. Our model incorporates two essential structural features—heterogeneity in hyperdegree distributions and the inter-order hyperedge overlap [ 33 ] , a metric which quantifies correlations between different orders of interactions. By analytically deriving the epidemic threshold and disentangling the contributions of two- and three-body interactions, we reveal how these structural features jointly shape the onset of epidemic outbreak. Crucially, we show that high heterogeneity in hyperdegree distributions can trigger explosive contagion, and that overlap modulates the early-stage activation of higher-order spreading pathways. These predictions are validated through Gillespie simulations on synthetic and empirical hypergraphs, and are shown to extend to SIS dynamics. Altogether, our results reveal the structural and dynamical mechanisms of abrupt transitions in contagion processes in the presence of group interactions, and offer a unifying framework to study spreading processes in real-world systems. Modeling higher-order interactions. —We model a system with higher-order interactions as a hypergraph H=(\mathcal{N},\mathcal{E}) , where \mathcal{N} is the set of N=|\mathcal{N}| nodes that interact via E=|\mathcal{E}| hyperedges, i.e., groups of nodes. Each hyperedge e\in\mathcal{E} , a subset of \mathcal{N} , can be characterized by its order m=|e|-1 , with m=1 representing pairwise interactions, m=2 corresponding to group interactions of three nodes, etc. We call \mathcal{E}_{m} the set of hyperedges of order m . Its counterpart is k_{m} , the generalized degree of order m , also called k -hyperdegree, denoting the number of m -hyperedges connected to a node [ 38 ] . We call P(k_{m}) their probability distribution, whose first and second moment, \langle k_{m}\rangle and \langle k_{m}^{2}\rangle , can thus be used to jointly quantify the mean connectivity and hyperdegree heterogeneity of higher-order networks. Notice, however, that these distributions alone provide no information about the microscopic arrangement of hyperedges or the correlations among different interaction orders. To quantify such correlations, we need to assess the extent to which a given structure adheres to or deviates from the inclusion property of simplicial complexes [ 39 ] . Calling \mathcal{F}(\mathcal{E}_{n}) the set of m -cliques within n -hyperedges, and given two orders of interaction m and n ( m<n ), the inter-order hyperedge overlap is given by [ 33 ] \alpha_{m,n}=\frac{\big|\mathcal{E}_{m}\cap\mathcal{F}(\mathcal{E}_{n})\big|}{\big|\mathcal{F}(\mathcal{E}_{n})\big|}, (1) where the numerator counts the m -cliques within n -hyperedges that are also m -hyperedges, normalized by the total number of m -cliques in n -hyperedges. This yields \alpha_{m,n}\in[0,1] , with \alpha_{m,n}=0 indicating no overlap, and \alpha_{m,n}=1 maximum overlap—when all m -cliques in n -hyperedges are also m -hyperedges. By construction, \alpha_{m,n}=0 for m>n . Higher-order networks constructed from empirical data exhibit a broad diversity of \alpha_{m,n} , as shown in SM. Figure 1: Group-based compartmental modeling. (a) Graphical representation of the model for M=2 . A test node u is connected to four infectious (red) and two susceptible (blue) nodes via two 1-hyperedges and two 2-hyperedges. Arrows show the different channels of infection through the rate parameters in Eq. ( 4 ). (b) Final epidemic size ( \mathcal{R}_{\infty} ) as a function of infectivity \lambda_{1} , with \lambda_{2}=3 . Model results (lines) are compared with Gillespie simulations (markers) on Scale-free and an Erdős-Rényi simplicial complexes, respectively with 2{,}000 and 10{,}000 nodes (see characteristics in Table 1 ). Group-based compartmental modeling.— To understand how correlations between different orders of interactions affect the onset and outcome of outbreaks, we propose a mathematical framework that explicitly includes the inter-order overlap of Eq. ( 1 ) as a free parameter. To this end, we consider the edge-based compartmental modeling (GBCM) approach for the SIR process [ 40 , 41 ] defined in [ 42 , 43 ] , generalizing it to capture infection dynamics within groups of different orders. Each order m=1,2,\ldots,M is associated with an infection rate \beta_{m} governing the rate at which a susceptible node is infected via a “contagious” m -hyperedge—defined as one in which all other m nodes are infectious.

Read the complete text