Detection of communities with Naming Game-based methods

Area of Science:

• Network Science
• Computational Social Science
• Sociophysics

Background:

• Complex networks exhibit emergent community structures based on agent interactions.
• Linguistic dynamics, particularly agreement and disagreement, are hypothesized to drive social network community formation.
• Existing community detection methods often rely solely on network topology.

Purpose of the Study:

• To demonstrate how social communication dynamics can reveal community structures in networks.
• To explore the emergent properties of community detection through opinion exchange.
• To investigate the impact of social features like trust, uncertainty, and opinion preference on community formation.

Main Methods:

• A computational model based on the Naming Game was developed.
• The model incorporates social features: trust, uncertainty, and opinion preference, which evolve through agent communication.
• The model was evaluated for its ability to detect both non-overlapping and overlapping communities.

Main Results:

• The addition of social features individually improved community detection capabilities.
• Evolving trust and uncertainty values provided insights into node and edge roles within the network.
• The model achieved accuracy comparable to specialized topological community detection algorithms.

Conclusions:

• Social communication dynamics, including trust and opinion preference, are effective in emergent community detection.
• The developed model offers a novel approach to understanding community structure formation in social networks.
• The model's ability to identify both simple and overlapping communities highlights its versatility.