[Paper Review] When group level is different from the population level: an adaptive network with the Deffuant model

Floriana Gargiulo, Sylvie Huet|arXiv (Cornell University)|Feb 9, 2010

Opinion Dynamics and Social Influence37 references3 citations

TL;DR

This paper proposes an adaptive network model that couples the Deffuant bounded confidence opinion dynamics with group formation, where individuals switch groups if their opinion differs significantly from their group's average. Surprisingly, the model achieves total consensus at lower confidence thresholds than the standard Deffuant model, due to group-level homogenization and preferential attachment to larger groups, leading to dynamic group size shifts and persistent opinion clusters at the population level despite internal consensus within groups.

ABSTRACT

We propose a model coupling the classical opinion dynamics of the bounded confidence model, proposed by Deffuant et al., with an adaptive network forming a community or group structure. At each step, an individual can decide if it changes groups or interact on its opinion with one of its internal or external neighbour. If it decides to look at the group level, it changes groups if its opinion is far from the average of its group from more than a threshold. If it is the case, it joins the group which has proportionally the closest average opinion from its. If it decides to interact with one of its neighbour, it becomes closer in opinion to it when its opinion and the one of the selected-to-interact neighbour are less distant from the threshold. From the study of this coupled model, we discover some surprising behaviours compared to the known behaviour of the Deffuant bounded confidence model(BC): The coupled model exhibits a total consensus for an threshold value lower than the BC model; the distribution of sizes of the groups changes: some groups become larger while other decrease in size, sometimes until containing only one individual; from the point of view of the groups, the consensus remains for a large set of threshold values while, looking at the population level, there are a lot of opinion clusters.

Motivation & Objective

• To investigate how group-level social structures influence opinion dynamics in adaptive networks.
• To examine the conditions under which consensus emerges at lower confidence thresholds than in the classical Deffuant model.
• To analyze the evolution of group size distribution and the role of preferential attachment in group formation.
• To understand the disconnect between group-level consensus and population-level opinion clustering.
• To explore the implications of group membership on individual mobility and opinion convergence.

Proposed method

• Individuals interact with neighbors based on the Deffuant bounded confidence model, updating opinions only if their difference is within a confidence threshold ε.
• At each time step, individuals assess whether to interact (opinion adjustment) or switch groups (network adaptation) based on their opinion's distance from their group's average.
• Group switching occurs if an individual's opinion differs from its group's average by more than ε; the individual joins the group with the closest average opinion.
• The network is adaptive: group membership evolves dynamically based on opinion homogeneity and perceived group similarity.
• The model uses a static network structure with intra-group and inter-group links, where group identity is defined by the majority of internal connections.
• Simulations track opinion distribution, group size evolution, and consensus formation across varying ε values.

Experimental results

Research questions

• RQ1 How does the introduction of adaptive group formation alter the consensus threshold in the Deffuant model?
• RQ2 What mechanisms drive the observed shift in group size distribution, including the emergence of singleton groups?
• RQ3 Why does the model exhibit total consensus at lower ε values than the standard Deffuant model?
• RQ4 To what extent does group-level consensus mask persistent opinion clustering at the population level?
• RQ5 How does preferential attachment to larger groups influence the stability and evolution of group structures?

Key findings

• The model achieves total consensus at ε = 0.11, which is lower than the threshold required in the standard Deffuant model, indicating that group-level dynamics suppress isolated opinion clusters.
• At ε = 0.11, all groups become opinion-homogeneous, with each group containing only individuals whose opinions are very close to the group average.
• Group size distribution becomes highly heterogeneous: some groups grow significantly larger while others shrink to a single individual, driven by preferential attachment to larger groups.
• For ε values between 0.11 and 0.2, group hierarchies and sizes remain stable, indicating a robust regime of group structure formation.
• Despite internal consensus within groups, the population-level opinion distribution remains fragmented into multiple clusters, especially at higher ε values.
• The model reveals a decoupling between group-level consensus and population-level opinion diversity, suggesting that group identity can mask broader societal polarization.