Source-linked AI summary
The Structure of Online Social Networks Mirror Those in the Offline World
R. I. M. Dunbar, Valerio Arnaboldi, Marco Conti, Andrea Passarella
TL;DR
The paper asks whether online social networks preserve the layered organization observed in offline face-to-face networks despite potentially weaker time constraints. It analyzes reciprocated communication in Facebook and Twitter ego networks, finding similarly scaled layers and an additional innermost layer at ~1.5 alters.
Problem
The study asks whether online networks still exhibit the layered structuring found in offline face-to-face networks despite potentially bypassing some time constraints.
Method
The authors analyze ego networks built from reciprocated Facebook and Twitter communication, using clustering to identify and size relationship layers.
Results
~1.5 alters form an additional innermost layer, while online layer scaling and sizes closely match offline face-to-face networks.
Takeaways & Limitations
Online communities have structural characteristics very similar to offline face-to-face networks.
Takeaways & Limitations
Facebook data reconstruction makes internal-layer estimates more accurate and external-layer estimates less precise.
Abstract
from arXiv · showhide
We use data on frequencies of bi-directional posts to define edges (or relationships) in two Facebook datasets and a Twitter dataset and use these to create ego-centric social networks. We explore the internal structure of these networks to determine whether they have the same kind of layered structure as has been found in offline face-to-face networks (which have a distinctively scaled structure with successively inclusive layers at 5, 15, 50 and 150 alters). The two Facebook datasets are best described by a four-layer structure and the Twitter dataset by a five-layer structure. The absolute sizes of these layers and the mean frequencies of contact with alters within each layer match very closely the observed values from offline networks. In addition, all three datasets reveal the existence of an innermost network layer at ~1.5 alters. Our analyses thus confirm the existence of the layered structure of ego-centric social networks with a very much larger sample (in total, >185,000 egos) than those previously used to describe them, as well as identifying the existence of an additional network layer whose existence was only hypothesised in offline social networks. In addition, our analyses indicate that online communities have very similar structural characteristics to offline face-to-face networks.
1. Introduction
The paper asks whether online social networks retain the layered structure of offline face-to-face networks despite potentially weaker time constraints. It examines Facebook and Twitter ego networks using reciprocated communication to identify and size these layers.
- Motivation: Online communication raises questions about constraints on social-network size and relationship structure.The paper focuses on whether digital technologies alter limits observed in offline networks.
- Offline baseline: Time and cognitive constraints are proposed to limit relationship investment and produce layered social networks.Individuals may invest social or emotional capital thickly in fewer alters or thinly in more.
- Research question: Relationships are identified from reciprocated online traffic, and algorithms search for layered patterns and estimate layer sizes.The analysis examines whether online ego networks resemble offline ego networks and, if so, how large their layers are.
- Offline baseline: Offline ego networks show successive layers near 5, 15, 50 and 150 alters, with scaling ratio about 3.These layers differ in contact frequency and emotional closeness.
- Research question: The study tests whether online networks exhibit comparable structuring despite internet communication potentially bypassing some face-to-face time constraints.It analyzes two Facebook datasets and one Twitter dataset.
2. Methods
The study constructs ego networks from Facebook and Twitter interaction data, estimates relationship contact frequency, and filters for socially active users and meaningful ties. Facebook data require reconstruction and sampling assumptions because privacy settings leave some interactions unobserved.
- Data: Two Facebook datasets and one Twitter dataset provide online interaction data for ego-network analysis.The Facebook data include friendship and wall interactions; the Twitter data include exchanges among users.
- Relationship definition: Active relationships have at least one interaction in the temporal windows used to estimate contact frequency.Differences in interaction counts across windows are interpreted as relationship intimacy.
- Filtering: 130,338 egos and 5,289,910 active edges remain after selecting users averaging more than 10 interactions per month.Inactive profiles are discarded to focus on socially active people.
- Filtering: Relationships with contact frequency above one message per year are retained to exclude ties receiving minimal time and cognitive investment.This threshold targets the active network rather than its more external layers.
3. Results
The online ego networks show clustered, approximately threefold-scaled layers that broadly match offline network structure. Facebook is best represented by four layers and Twitter can support an additional layer, while all datasets reveal a new innermost layer.
- Cluster structure: 4.35 clusters on average for Facebook dataset #1 and 4.10 for Facebook dataset #2, with medians of 4.Twitter has a modal optimum near 4 but a median of 5 and a long right tail.
- Cluster structure: Facebook is analyzed with k=4, whereas Twitter is evaluated with k=4 and k=5.Twitter’s k=5 solution matches offline layer sizes better than k=4 under k-means, while DBSCAN favors k=4.
- Robustness: K-means and DBSCAN produce very similar layer sizes, with the final layer differing by at most about 3 alters.This supports k-means identification of clusters despite its simplicity.
- Contact frequency: Facebook contact thresholds range from approximately every five days in layer 0 to every six months in layer 3.These frequencies are compatible with face-to-face network observations.
- Contact frequency: Twitter contact thresholds range from every one or two days in layer 0 to two or three times yearly in layer 4.Twitter frequencies are higher than Facebook frequencies, especially in the innermost layers.
4. Discussion
Across three large online datasets, ego-centric networks reproduce the layered structure of offline face-to-face networks, including comparable layer sizes, scaling ratios, and contact frequencies. The analyses also reveal a robust innermost layer of approximately 1.5 individuals, while Facebook networks lack the usual outermost layer under the available data.
- Layered structure: Three online datasets confirm layered ego-centric networks with relative sizes and scaling ratios comparable to offline networks.The evidence is based on more than 185,000 egos across three different datasets.
- Scope and limitations: Online ego networks are smaller than conventional offline networks, particularly in Facebook, where the outermost layer is entirely absent.The missing weak-tie layer may reflect insufficient information about very low-frequency relationships in the Facebook data.
- Layered structure: Facebook datasets reproduce the conventional cumulative layers associated with 5, 15, and 50 individuals, while Twitter also shows the 150-person outer layer.The paper describes Facebook as lacking the external layer corresponding to roughly 150 alters.
- Innermost layer: ~1.5 individuals form an innermost layer visible in all three datasets and scaling consistently with the layers outside it.These alters are contacted at least once every five days on Facebook and every other day on Twitter, indicating very high emotional investment.
- Contact frequencies: Mean contact rates in online layers closely match offline networks, especially for Facebook, although Twitter rates are generally higher.Twitter frequencies are about double the corresponding Facebook and face-to-face rates while retaining similar ratios across layers.
- Interpretation: The authors interpret the similarities as evidence that online environments map closely onto everyday offline networks or become integrated into personal networks.They frame these findings as sociological similarities rather than as a claim that online communication causes offline network structure.
Acknowledments
The research is supported by a European Research Council Advanced grant.
- European Research Council Advanced grant support is acknowledged.
- The acknowledged funding comes from the European Research Council.
- The grant is identified as an Advanced grant.