Dynamics in online social networks

Przemyslaw A. Grabowicz, José J. Ramasco, Víctor M. EguíluzarXiv 2012 · DOI · OpenReview

TL;DR

This paper provides a comprehensive overview of the structure, dynamics, and activity within online social networks, drawing parallels and distinctions with offline social behavior. It highlights how network science principles can be applied to understand these complex digital ecosystems, offering insights into group growth, information diffusion, and the validity of sociological theories in online contexts.

What Problem Does This Paper Address?

The paper addresses the growing importance of online social networks as platforms for human interaction and the need to understand their underlying structure and dynamics. While many online platforms exist, they all rely on an underlying social network that facilitates connections and activities. The challenge is to apply established network science methodologies, alongside sociological and computer science perspectives, to analyze these complex, large-scale systems and to determine if theories developed for offline social behavior hold true in the digital realm. Specifically, it seeks to understand how these networks grow, how information spreads through them, and how user activity relates to declared social ties.

Methods

The paper reviews and discusses various methods and models used in the study of online social networks. These include:

  • Network Science Fundamentals: Analysis of degree distributions (often heavy-tailed, like power-laws or log-normals), clustering coefficients (global and local), and the small-world phenomenon (low average shortest-path distance).
  • Growth Models: Discussion of preferential attachment models (like Barabási-Albert) and heterogeneous growth models incorporating "fitness" or hidden variables. The paper also highlights triangle-closing models as particularly effective for reproducing the high clustering observed in social networks.
  • Community Detection: Mention of algorithms used to identify densely connected parts of the network, as well as the use of user-declared groups.
  • Activity Analysis: Comparison of networks built from declared online friendships versus those derived from user interactions (e.g., mentions, retweets).
  • Testing Sociological Theories: Empirical examination of theories like Granovetter's "strength of weak ties" and Burt's work on brokerage and closure, using online data to see if they apply.

Datasets

The paper references studies that utilize various datasets, including:

  • Flickr: Used for studying group growth dynamics.
  • Twitter: Used for analyzing internal activity, link types, and information diffusion.
  • Mobile phone call data: Used in some studies to test social tie strength.
  • Email networks: Mentioned as an early example for studying social tie definitions.
  • Facebook: Mentioned in the context of a large field experiment on information propagation.

The paper does not introduce a new benchmark dataset but rather synthesizes findings from existing research that used these diverse sources.

Key Insights and Ideas

  • Online vs. Offline Parallels: Many structural properties of offline social networks, such as heavy-tailed degree distributions and the small-world effect, are also observed in online networks.
  • The Importance of Clustering: High clustering (transitivity) is a defining feature of social networks, meaning friends of friends are likely to be friends themselves. This is crucial for community formation.
  • Growth Mechanisms: While preferential attachment is a foundational model, more sophisticated models incorporating heterogeneity (like "fitness") and triangle-closing mechanisms are better at capturing the observed growth patterns and structural properties of online social networks.
  • Activity vs. Declaration: The network of actual user interactions (e.g., mentions, retweets) often differs significantly from the declared network of online friends, with lower degrees and more transient connections in the activity network.
  • Sociological Theories Hold Up (Mostly): Theories like the "strength of weak ties" appear to have validity in online social networks. Weak ties (less frequent, less intimate) are often found to be more effective for information diffusion across different groups, while strong ties are more internal to communities.
  • Tradeoffs in Information Diffusion: The efficiency of information spread is influenced by a tradeoff between the diversity of information sources and the bandwidth of communication ties.

Key Results

  • Group Growth: Heterogeneous linear growth models, which incorporate intrinsic "fitness" for groups, provide a more realistic representation of group growth in platforms like Flickr compared to simpler models like the Simon model. These models better capture the distribution of group sizes and the correlation between group size and age.
  • Link Types and Activity: In Twitter, links with mentions are more likely to be internal to communities, while links between groups or intermediary links are more important for retweets, suggesting a role in information propagation across diverse user bases.
  • Tie Strength and Diffusion: Studies on Facebook and Twitter suggest that users are significantly more likely to re-broadcast information from friends, with weaker ties potentially being more efficient for information flow due to their role in bridging communities.
  • Community Structure: The presence of communities is a robust finding, and interactions tend to be more concentrated within these communities.

Limitations and Future Work

  • Understanding Fitness: While heterogeneous models show the importance of intrinsic "fitness" for group growth, the specific factors that determine this fitness in online social networks remain an open question.
  • Validity of Online Data: There's an ongoing discussion about the extent to which online interactions accurately reflect real-world social activity and the limitations of using online networks as indicators. Further investigation is needed.
  • Network Densification: The phenomenon of network densification over time is mentioned as an area requiring further study.
  • Future Research: The authors express optimism that the field of online social network research will continue to be active, with ongoing observations and experiments expected to provide deeper quantitative descriptions of social behaviors. They anticipate that this research will continue to bridge the gap between sociological theories and observed online phenomena.

Summary updated

Abstract

An increasing number of today's social interactions occurs using online social media as communication channels. Some online social networks have become extremely popular in the last decade. They differ among themselves in the character of the service they provide to online users. For instance, Facebook can be seen mainly as a platform for keeping in touch with close friends and relatives, Twitter is used to propagate and receive news, LinkedIn facilitates the maintenance of professional contacts, Flickr gathers amateurs and professionals of photography, etc. Albeit different, all these online platforms share an ingredient that pervades all their applications. There exists an underlying social network that allows their users to keep in touch with each other and helps to engage them in common activities or interactions leading to a better fulfillment of the service's purposes. This is the reason why these platforms share a good number of functionalities, e.g., personal communication channels, broadcasted status updates, easy one-step information sharing, news feeds exposing broadcasted content, etc. As a result, online social networks are an interesting field to study an online social behavior that seems to be generic among the different online services. Since at the bottom of these services lays a network of declared relations and the basic interactions in these platforms tend to be pairwise, a natural methodology for studying these systems is provided by network science. In this chapter we describe some of the results of research studies on the structure, dynamics and social activity in online social networks. We present them in the interdisciplinary context of network science, sociological studies and computer science.
BibTeX
@misc{grabowicz2012dynamics,
  author = {Przemyslaw A. Grabowicz and José J. Ramasco and Víctor M. Eguíluz},
  title = {{Dynamics in online social networks}},
  year = {2012},
  eprint = {1210.0808},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1210.0808},
}

Loading related content...

Figure