Consensus formation in opinion dynamics with online and offline interactions at complex networks

Zhaogang Ding, Yucheng Dong, Gang Kou, Iván Palomares, Shui YuInternational Journal of Modern Physics C 2018 · DOI

TL;DR

This paper investigates how the differing speeds of information exchange in online versus offline social interactions affect the formation of consensus within complex networks. They find that a significant difference in update speeds (asynchronism) hinders consensus, especially when online participants dominate, offering insights for understanding real-world opinion dynamics and policy communication.

What Problem Does This Paper Address?

In today's interconnected world, individuals receive and share information through both online platforms (like social media) and traditional offline interactions (face-to-face conversations). These two modes of communication operate at vastly different speeds. Online interactions are typically much faster, allowing for rapid dissemination of information and opinion updates, while offline interactions are slower. This disparity in speed, termed "asynchronism," is a crucial factor in how opinions evolve and whether a group can reach a shared consensus. Existing models of opinion dynamics often assume a synchronized update process, where all agents update their opinions simultaneously. This paper aims to bridge this gap by exploring how this inherent asynchronism between online and offline interactions influences consensus formation in complex social networks. Understanding this is vital for fields ranging from social science to marketing and even political science, as it impacts how information spreads and how collective decisions are made.

Methods

The researchers propose an asynchronous opinion dynamics model built upon the Hegselmann-Krause (HK) bounded confidence model. In this model, agents are divided into two types: online and offline. The key innovation is the introduction of a parameter, T, which dictates the asynchronous update rate. When T = 1, the system is fully synchronized. When T > 1, offline agents update their opinions at a slower pace compared to online agents. Specifically, offline agents only update their opinions at time intervals that are multiples of T, while online agents update at every time step.

The model simulates opinion dynamics on three types of complex networks: Erdős-Rényi (ER) random networks, Barabási-Albert (BA) scale-free networks, and Watts-Strogatz (WS) small-world networks. They analyze two primary metrics:

  • Consensus Ratio: The proportion of simulation samples where all agents reach a single opinion cluster (i.e., a consensus).
  • Consensus Threshold (εc): The minimum confidence level required for almost all agents to reach a consensus.

Extensive computational simulations were performed across various network structures, confidence levels (ε), degrees of asynchronism (T), and proportions of online agents (r).

Datasets

Not discussed in the paper. The study relies on simulated data generated from their proposed opinion dynamics model on various network structures.

Key Insights and Ideas

  • Asynchronism is a double-edged sword for consensus: A low degree of asynchronism (small T) actually enhances consensus formation. This is because it allows for a more coordinated, albeit slightly delayed, exchange of opinions, facilitating agreement.
  • High asynchronism hinders consensus: Conversely, a high degree of asynchronism (large T) significantly reduces the likelihood of reaching consensus. The growing gap in update speeds between online and offline agents creates divergence.
  • Online dominance amplifies effects: The impact of asynchronism is amplified when the proportion of online agents (r) is high. This is intuitive, as faster online updates can lead to rapid opinion clustering among online agents, potentially leaving offline agents behind.
  • Network structure matters, but the trend persists: While the specific consensus ratios and thresholds vary across ER, BA, and WS networks, the general trends related to asynchronism and the proportion of online agents remain consistent.
  • Consensus threshold is sensitive to asynchronism: The required confidence level (εc) to achieve consensus is significantly affected by the degree of asynchronism. Low asynchronism can lower the threshold, while high asynchronism raises it.

Key Results

  • Low asynchronism (T=2) improves consensus: For small values of T, the consensus ratio generally increases compared to full synchronization (T=1).
  • High asynchronism (T=100) reduces consensus: As T increases, the consensus ratio tends to decrease, especially for higher proportions of online agents.
  • Online proportion's role: When asynchronism is high, a larger proportion of online agents (r) leads to a lower consensus ratio, up to a point (around 80% online agents), after which the ratio may increase moderately.
  • Consensus threshold dynamics: The consensus threshold (εc) generally decreases with increasing average network degree. For WS networks, a higher cluster coefficient leads to a lower consensus threshold. For high asynchronism, a larger proportion of online agents leads to a higher consensus threshold.

Limitations and Future Work

  • Simplified model of reality: The paper acknowledges that the model simplifies real-world interactions. For instance, it assumes a fixed update speed ratio (T) and a binary classification of agents (online/offline). Future work could explore more nuanced models of interaction speeds and agent types.
  • Limited network types: While three common network types were studied, exploring other complex network structures could provide broader insights.
  • Focus on specific metrics: The study primarily focuses on consensus ratio and threshold. Future research could investigate other emergent properties of opinion dynamics under asynchronism, such as polarization or opinion diversity.
  • Real-world data validation: While the findings offer theoretical insights, direct validation with real-world data from online and offline communication platforms would strengthen the conclusions.
  • Governmental policy implications: The authors suggest that governments can leverage low degrees of asynchronism to promote consensus on policies, while high degrees might necessitate more support for offline interactions to bridge opinion gaps. This practical application could be further explored.

Summary updated

Abstract

Nowadays, with the development of information communication technology and Internet, more and more people receive information and exchange their opinions with others via online environments (e.g. Twitter, Facebook, Weibo, and WeChat). According to eMarketer Report [Worldwide Internet and Mobile Users: eMarketer’s Updated Estimates and Forecast for 2015–2020 (eMarketer Report). Published October 11, 2016, https://www.emarketer.com/Report/Worldwide-Internet-Mobile-Users-eMarketers-Updated-Estimates-Forecast-20152020/2001897 ).], by the end of 2016, more than 3.2 billion individuals worldwide will use the Internet regularly, accounting for nearly 45% of the world population. By contrast, the other half of the global population still obtain information and regularly exchange their opinions in a more traditional way (e.g. face to face). Generally, the speed at which information spreads and opinions are exchanged and updated in an online environment is much faster than in an offline environment. This paper focuses on jointly investigating the challenge of consensus formation in opinion dynamics with online and offline interactions. Without loss of generality, we assume the speed at which information spreads and opinions are exchanged and updated in an online environment is [Formula: see text] times as fast as in an offline environment. We demonstrate that the update speed ratio in mixed online and offline environments (i.e. [Formula: see text]) strongly impacts the consensus formation at complex networks: a large update speed ratio of online and offline environments (i.e. [Formula: see text]) makes it difficult for all agents to reach consensus in opinion dynamics. Furthermore, these effects are often further intensified as the number of online participating agents increases.
BibTeX
@article{ding2018consensus,
  author = {Zhaogang Ding and Yucheng Dong and Gang Kou and Iván Palomares and Shui Yu},
  title = {{Consensus formation in opinion dynamics with online and offline interactions at complex networks}},
  journal = {International Journal of Modern Physics C},
  year = {2018},
  doi = {10.1142/s0129183118500468},
}
Figure