[Defense of thesis] 28/11/2024 - Emmanuel KRAVITZCH : "Opinion dynamics in coevolution with adaptive networks" (UPR LIA)

Emmanuel KRAVITZCH will defend his thesis on 28 November 2024 on the theme: "Dynamics of opinion in coevolution with adaptive networks".

Laboratory

UPR 4128 LIA - Avignon Computing Laboratory

Composition of the jury

M. Vineeth, S VARMA	CRAN, University of Nancy	Thesis co-supervisor
Mrs Armelle BRUN	University of Lorraine	Rapporteur
Ms Elena VALCHER	University of Padua	Rapporteur
Yezekael HAYEL	University of Avignon, Computer Science Laboratory (LIA, CERI)	Thesis supervisor
Mr Antoine, Olivier BERTHET	CentraleSupélec, Université Paris-Saclay	Thesis co-director
Mr Piet VAN MIEGHEM	Delft University	Examiner
Mr Walid BEN-AMEUR	Télécom Sud-Paris	Examiner
Pierre-Henri MORAND	University of Avignon	Guest

Keywords : evolutionary networks, stochastic analysis, modelling, complex systems

Summary

This thesis is part of the broader movement towards the analysis of complex systems and networks. My doctoral work falls within the framework of coevolution, a sub-field of network analysis. The common thread running through this thesis has been the adaptive nature of the communication network in response to the state profile of the agents making up the network nodes. We call it co-evolutionary because the two dynamics-namely, the dynamics of the graph and the dynamics of the nodes-evolve together and influence each other. The central objective of the thesis was to design analytically affordable co-evolutionary models that exhibit interesting behaviours. The thesis manuscript therefore consists of two parts. In the first part, we discuss Voter models on adaptive networks. The first chapter of the first part (chap. 2) presents a well-known spin model, the Votant model, outlines some general ideas and motivates the adaptive extension - where the graph is variable. In the second chapter of the first part (chap. 3), we deal with a specific situation: a single agent confronted with two opposing sources of influence. This might represent, for example, a citizen during a political campaign. This case is analysed using a probabilistic technique known as the fluid limit. Finally, the last chapter of the first part (chapter 4) provides a general analysis of a co-evolutionary model using a mean-field approximation. The results are discussed, refined with an alternative approximation, and finally illuminated by numerous numerical simulations. The second part (part II) focuses on another classical framework for the exchange of opinions between agents: the bounded trust model. After a brief historical overview of this type of model (chap. 5), an original multi-layer model is proposed (chap. 6). In this model, agents can interact via two different communication channels, which leads to results that are distinct from classical bounded trust models. Finally, the last chapter of the second part (chapter 7) revisits the bounded trust model by incorporating an evolution equation for the agents' "self-confidence". Finally, the concluding chapter (chap. 8) takes a critical look at my work and suggests some very concrete avenues for future research.