Reducibility of higher-order to pairwise interactions: Social impact models on hypergraphs

Jaume Llabrés, Raúl Toral, Maxi San Miguel, Federico Vázquez · 2026

Read Reducibility of higher-order to pairwise interactions: Social impact models on hypergraphs by Jaume Llabrés, Raúl Toral, Maxi San Miguel, Federico Vázquez (2026) 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-2601-05169.txt.

Introduction

Jaume Llabrés and 3 co-authors show that many higher-order social models can be rewritten exactly as simpler pairwise models in a 2026 paper. Social impact models on hypergraphs let groups, not just pairs, influence each other, which seems to demand new mathematics. The authors prove a general class of such models reduces to a pairwise model on a weighted network that behaves in exactly the same way. The reduction preserves the microscopic probabilities of nodes changing state.

As a special case they introduce hypergraph-voter models, computing the projected weights both on paper and by simulation for several hypergraph families. Simulations compare the ordering dynamics of both versions. An exact reduction means researchers can study group influence with the mature tools built for ordinary networks.

Catalog record

CREATORJaume Llabrés, Raúl Toral, Maxi San Miguel, Federico Vázquez
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE98 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-05169

Browse the full library · About the Tower of Babel

Full text

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

Reducibility of higher-order to pairwise interactions: Social impact models on hypergraphs Jaume Llabrés, Raúl Toral, Maxi San Miguel, Federico Vázquez Abstract We show that a general class of social impact models with higher-order interactions on hypergraphs can be exactly reduced to an equivalent model with pairwise interactions on a weighted projected network. This reduction is made by a mapping that preserves the microscopic probabilities of changing the state of the nodes. As a particular case, we introduce hypergraph-voter models, for which we compute the weights of the projected network both analytically and numerically across several hypergraph ensembles, and we characterize their ordering dynamics through simulations of both higher-order and reduced dynamics. For a linear social impact function ( hypergraph-linear voter model ) the weights of the projected network are static, allowing us to develop a pair approximation that describes with accuracy the time evolution of macroscopic observables, which turn out to be independent of those weights. The macroscopic dynamics is thus equivalent to that of the standard voter model on the unweighted projected network. For a power-law social impact function ( hypergraph-nonlinear voter model ) the weights of the projected network depend on the instantaneous system configuration. Nevertheless, the nonlinear voter model on the unweighted projected network still reproduces the main macroscopic trends for well connected hypergraphs. I Introduction Over the last decade, it has been emphasized that many biological, social, technological and physical systems cannot be fully described by pairwise (PW) dyadic interactions alone. In many cases, the fundamental units of interaction are often polyadic , involving three or more entities. Therefore, relying solely on pairwise interactions may thus obscure important features of social dynamics 33 ; 16 ; 6 ; 49 ; 5 . Hypergraphs and simplicial complexes provide natural mathematical frameworks to encode such higher-order (HO) interactions and to explore the rich collective phenomena that they generate 7 . Unlike standard complex networks—where nodes represent agents and edges represent dyadic interactions—hypergraphs allow hyperedges containing multiple nodes, each hyperedge corresponding to an interacting group. Recent reviews have shown that incorporating hyperedges in the description of a system profoundly alters both its structural and dynamical properties, while highlighting key open theoretical challenges 6 ; 4 ; 19 . Simplicial-complex models have also been used to describe tightly bound groups, high-dimensional social structures 30 , and the ways in which group members acquire and accumulate information 25 . However, not all group or many-agent interactions require an explicit hypergraph representation. Certain social processes—such as multiple exposures inducing social conformity 3 —can be adequately captured within a standard complex networks framework. For instance, phenomena of complex contagion 14 ; 13 , including adoption of innovations or propagation of rumors, are often described by models inspired in the threshold model by Granovetter 24 ; 55 in which, in a complex network representation, adoption requires a minimum number of neighboring adopters. It is known that complex contagion models often exhibit a cascade mechanism leading to a discontinuous first-order transition 55 ; 23 , in contrast to the typically continuous transitions found in dyadic epidemic models 45 . Yet similar discontinuous transitions have also been attributed to HO interactions in hypergraphs and simplicial complexes 28 ; 16 . A second example is the nonlinear voter model on a complex network 48 ; 32 ; 37 , in which the update of the state on an agent depends on the joint configuration of all its neighbors, effectively capturing the simultaneous influence of group interactions. In fact, the presence of a group interaction is determined not by the network’s topology, but rather by the dynamics defined on the network: In the nonlinear voter model group interactions arise from update rules defined on an otherwise dyadic network. In the Statistical Physics treatment of interacting particles, interactions are typically assumed to be pairwise, but this does not preclude the emergence of three-body or higher-order collisions, as captured by the Ursell–Mayer cluster expansion 27 ; 46 . The diagrams of this expansion resemble hyperedges in modern HO interactions literature. In the virial expansion of systems with purely pairwise interaction potentials, the second-order coefficient reflects two-body collisions, while higher-order coefficients correspond to collisions among larger sets of particles. This illustrates that many-body effects—analogous to HO interactions—can arise even within fundamentally pairwise interaction frameworks. It is in the context of the above discussion that the central questions that motivate the present work arise. What is exactly meant or implied by a “higher-order interaction”? Does it correspond to a genuine group interaction, encoded by an explicit three-body or higher-order potential, or can it emerge from combinations of pairwise interactions? When is a hypergraph representation necessary? Under what conditions can HO interactions be reduced to effective dyadic interactions? In fact, the reducibility of HO interactions to effective PW interactions has been already addressed at different levels. A first microscopic approach studies when HO interactions on a hypergraph can be rewritten in terms of PW interactions 44 ; 49 . A key result of these studies, restricted to continuous state variables of the nodes ( s\in[0,1] ), is that linear interactions on hypergraphs can be always mapped onto PW interactions on a weighted complex network. A second approach considers reducibility at a macroscopic level, identifying dyadic-based models that reproduce the same macroscopic equations than those derived from HO interactions on a hypergraph 42 . Related to this macroscopic reducibility, is the question of the minimal model capable of reproducing macroscopic data with given precision 56 ; 41 or that of dimensionality reduction 40 . In this work we examine the reducibility of HO interactions in social impact models 34 ; 32 , defined on hypergraphs. In these models, each agent is endowed with a binary state s=\pm 1 and updates its state through interactions with groups of other agents, represented by hyperedges. This interaction is characterized by a social impact function. Because each update depends on all agents within the hyperedge, these interactions are genuinely higher order. Our main result is that a general social impact model on an arbitrary hypergraph can be always reduced at a microscopic level to a model of PW interactions on a projected complex network with weighted links. This mapping is exact: the stochastic dynamics of HO interactions on the hypergraph is equivalent to a PW dynamics on the projected network, with link weights encoding the probability of interaction between node pairs and, in general, depending on the instantaneous state configuration. As an illustrative class of social impact models, we introduce hypergraph-voter models 31 , in which agents update their state according to the power q of the fraction of agents in the same hyperedge holding the opposite opinion. For the hypergraph-linear voter model ( q=1 ), the mapping yields a projected network with time-independent weights. At the macroscopic level, the hypergraph-linear voter model features the same evolution equations as the standard voter model on the projected network with unweighted links (VM) 1 ; 26 ; 35 ; 36 ; 52 ; 53 ; 51 ; 54 ; 12 , implying that macroscopic behavior can be captured by assuming homogeneous weights. By contrast, the hypergraph-nonlinear voter model ( q\neq 1 ) generates effective pairwise interactions on a projected network

Read the complete text