What's in a Niche? Migration Patterns in Online Communities

TL;DR

The paper investigates how users migrate among topic-based online communities and whether these migrations follow a directional gradient. It formalizes three gradients—user, informational, and referential—each yielding acyclic partial orders across communities. Empirical analysis on Reddit topics and Wikipedia shows that migrations tend toward smaller, less toxic, more linguistically distinctive communities, which supports specialization rather than radicalization. The authors validate findings with null models and simulations, and discuss implications for recommendations, moderation, and bot detection across platforms.

Abstract

Broad topics in online platforms represent a type of meso-scale between individual user-defined communities and the whole platform; they typically consist of related communities that address different facets of a shared topic. Users often engage with the topic by moving among the communities within a single category. We find that there are strong regularities in the aggregate pattern of user migration, in that the communities comprising a topic can be ordered in a partial order such that there is more migration in the direction defined by the partial order than against it. Ordered along this overall direction, we find that communities in aggregate become smaller, less toxic, and more linguistically distinctive, suggesting a picture consistent with specialization. We study directions defined not just in the movement of users but also by the movement of URLs and by the direction of mentions from one community to another; each of these produces a consistent direction, but the directions all differ from each other. We show how, collectively, these distinct trends help organize the structure of large online topics and compare our findings across both Reddit and Wikipedia and in simulations.

Figures (8)

• Figure 1: The user gradient for sports communities on Reddit, one of the topics we study, demonstrates how three of our measures (size, toxicity, and linguistic distinctiveness) are related to the aggregate progression of users along a gradient - communities become smaller, less toxic, and more distinct. Edges proceed downward. The size of nodes is a logarithmic scale of the size of communities. The inner color of the nodes represents the toxicity of the community, where darker nodes are more toxic. The outer color of the node represents linguistic distinctiveness, where darker edges represent less distinctive communities.
• Figure 2: Histogram of the asymmetry scores for all pairs of communities. Adjusting $k$ has little effect. Note that the average asymmetry of user directions is far more than would be expected with random movement (with a p-value of nearly 0), and random movement would produce an exponential distribution rather than a more uniform distribution.
• Figure 3: Small graphs representing the user gradient for politics and CS, where all edges move downwards. Note that not all communities in each category appear in these graphs. Those that do not, do not have strong enough user overlap with any other community ($<20$ users by our definition). For CS, we remove all edges of less than 0.75 asymmetry to remain acyclic, our only thresholding for the use gradient.
• Figure 4: Histograms of the asymmetry scores for all pairs of communities by URLs.
• Figure 5: Histograms of the asymmetry scores for all pairs of communities by mentions.

Theorems & Definitions (4)

• Definition 1: Thresholding
• Definition 2: Expanding
• Definition 3: Orientation & Asymmetry
• Definition 4: Gradient