The structure and dynamics of networks with higher order interactions

Physics Reports, Journal Year: 2023, Volume and Issue: 1018, P. 1 - 64

Published: May 1, 2023

Language: Английский

---

A Survey on Hypergraph Representation Learning

ACM Computing Surveys, Journal Year: 2023, Volume and Issue: 56(1), P. 1 - 38

Published: June 22, 2023

Hypergraphs have attracted increasing attention in recent years thanks to their flexibility naturally modeling a broad range of systems where high-order relationships exist among interacting parts. This survey reviews the newly born hypergraph representation learning problem, whose goal is learn function project objects—most commonly nodes—of an input hyper-network into latent space such that both structural and relational properties network can be encoded preserved. We provide thorough overview existing literature offer new taxonomy embedding methods by identifying three main families techniques, i.e., spectral, proximity-preserving, (deep) neural networks. For each family, we describe its characteristics our insights single yet flexible framework then discuss peculiarities individual methods, as well pros cons. review tasks, datasets, settings which embeddings are typically used. finally identify open challenges would inspire further research this field.

Language: Английский

---

Contagion dynamics on higher-order networks

Nature Reviews Physics, Journal Year: 2024, Volume and Issue: 6(8), P. 468 - 482

Published: July 5, 2024

Language: Английский

---

Influential groups for seeding and sustaining nonlinear contagion in heterogeneous hypergraphs

Communications Physics, Journal Year: 2022, Volume and Issue: 5(1)

Published: Jan. 17, 2022

Abstract Contagion phenomena are often the results of multibody interactions—such as superspreading events or social reinforcement—describable hypergraphs. We develop an approximate master equation framework to study contagions on hypergraphs with a heterogeneous structure in terms group size (hyperedge cardinality) and node membership (hyperdegree). By mapping interactions nonlinear infection rates, we demonstrate influence large groups two ways. First, characterize phase transition, which can be continuous discontinuous bistable regime. Our analytical expressions for critical tricritical points highlight first three moments distribution. also show that sizes contagion promote mesoscopic localization regime where is sustained by largest groups, thereby inhibiting bistability. Second, formulate optimal seeding problem hypergraph compare strategies: allocating seeds according properties. find that, when sufficiently nonlinear, more effective than individual hubs.

Language: Английский

---

Multistability, intermittency, and hybrid transitions in social contagion models on hypergraphs

Nature Communications, Journal Year: 2023, Volume and Issue: 14(1)

Published: March 13, 2023

Abstract Although ubiquitous, interactions in groups of individuals are not yet thoroughly studied. Frequently, single modeled as critical-mass dynamics, which is a widespread concept used only by academics but also politicians and the media. However, less explored questions how collection will behave their intersection might change dynamics. Here, we formulate this process binary-state dynamics on hypergraphs. We showed that our model has rich behavior beyond discontinuous transitions. Notably, have multistability intermittency. demonstrated phenomenology could be associated with community structures, where or intermittency controlling number size bridges between communities. Furthermore, provided evidence observed transitions hybrid. Our findings open new paths for research, ranging from physics, formal calculation quantities interest, to social sciences, experiments can designed.

Language: Английский

---

Higher-order temporal interactions promote the cooperation in the multiplayer snowdrift game

Science China Information Sciences, Journal Year: 2023, Volume and Issue: 66(12)

Published: Nov. 27, 2023

Language: Английский

---

A unified framework for simplicial Kuramoto models

Chaos An Interdisciplinary Journal of Nonlinear Science, Journal Year: 2024, Volume and Issue: 34(5)

Published: May 1, 2024

Simplicial Kuramoto models have emerged as a diverse and intriguing class of describing oscillators on simplices rather than nodes. In this paper, we present unified framework to describe different variants these models, categorized into three main groups: “simple” “Hodge-coupled” “order-coupled” (Dirac) models. Our is based topology discrete differential geometry, well gradient systems frustrations, permits systematic analysis their properties. We establish an equivalence between the simple simplicial model standard pairwise networks under condition manifoldness complex. Then, starting from notion synchronization derive bounds coupling strength necessary or sufficient for achieving it. For some variants, generalize results provide new ones, such controllability equilibrium solutions. Finally, explore potential application in reconstruction brain functional connectivity structural connectomes find that edge-based perform competitively even outperform complex extensions node-based

Language: Английский

---

The temporal dynamics of group interactions in higher-order social networks

Nature Communications, Journal Year: 2024, Volume and Issue: 15(1)

Published: Aug. 27, 2024

Representing social systems as networks, starting from the interactions between individuals, sheds light on mechanisms governing their dynamics. However, networks encode only pairwise interactions, while most occur among groups of requiring higher-order network representations. Despite recent interest in little is known about that govern formation and evolution groups, how people move groups. Here, we leverage empirical data children university students to study temporal dynamics at both individual group levels, characterising individuals navigate form disaggregate. We find robust patterns across contexts propose a dynamical model closely reproduces observations. These results represent further step understanding systems, open up research directions impact processes evolve top them. The structure many where human involve communities can be described by networks. authors hypergraph-based describes different sizes.

Language: Английский

---

Shaping new norms for AI

Philosophical Transactions of the Royal Society B Biological Sciences, Journal Year: 2024, Volume and Issue: 379(1897)

Published: Jan. 21, 2024

As artificial intelligence (AI) becomes increasingly integrated into our lives, the need for new norms is urgent. However, AI evolves at a much faster pace than characteristic time of norm formation, posing an unprecedented challenge to societies. This paper examines possible criticalities processes formation surrounding AI. It focuses on how can be established, rather what these should be. distinguishes different scenarios based centralization or decentralization process, analysing cases where are shaped by formal authorities informal institutions, emerge spontaneously in bottom-up fashion. On latter point, reports conversation with ChatGPT which LLM discusses some emerging it has observed. Far from seeking exhaustiveness, this article aims offer readers interpretive tools frame society’s response growing pervasiveness An outlook could influence future social emphasizes importance open societies anchor their deliberation process open, inclusive and transparent public discourse. part theme issue ‘Social change: drivers consequences’.

Language: Английский

---

Emergent social conventions and collective bias in LLM populations

Science Advances, Journal Year: 2025, Volume and Issue: 11(20)

Published: May 14, 2025

Social conventions are the backbone of social coordination, shaping how individuals form a group. As growing populations artificial intelligence (AI) agents communicate through natural language, fundamental question is whether they can bootstrap foundations society. Here, we present experimental results that demonstrate spontaneous emergence universally adopted in decentralized large language model (LLM) agents. We then show strong collective biases emerge during this process, even when exhibit no bias individually. Last, examine committed minority groups adversarial LLM drive change by imposing alternative on larger population. Our AI systems autonomously develop without explicit programming and have implications for designing align, remain aligned, with human values societal goals.

Language: Английский

---