Centralization Problem for Opinion Convergence in Decentralized Networks

TL;DR

This paper explores how to efficiently centralize control in decentralized opinion networks by selecting minimal influential agents, using new algorithms and validating their effectiveness on various networks.

Contribution

It introduces a novel framework for centralization in opinion dynamics, focusing on dominating set selection and proposing the prowling algorithm for optimal access unit placement.

Findings

The prowling algorithm effectively identifies minimal access units.

Our method outperforms existing benchmarks in real-world and synthetic networks.

Centralization can be achieved with fewer access units using the proposed approach.

Abstract

This paper aims to provide a new perspective on the interplay between decentralization -- a prevalent character of multi-agent systems -- and centralization, i.e., the task of imposing central control to meet system-level goals. In particular, in the context of networked opinion dynamic model, the paper proposes and discusses a framework for centralization. More precisely, a decentralized network consists of autonomous agents and their social structure that is unknown and dynamic. Centralization is a process of appointing agents in the network to act as access units who provide information and exert influence over their local surroundings. We discuss centralization for the DeGroot model of opinion dynamics, aiming to enforce opinion convergence using the minimum number of access units. We show that the key to the centralization process lies in selecting access units so that they form a dominating set. We then propose algorithms under a new local algorithmic framework, namely prowling, to accomplish this task. To validate our algorithm, we perform systematic experiments over both real-world and synthetic networks and verify that our algorithm outperforms benchmarks.