Mapping Social Media User Behaviors in Reciprocity Space

TL;DR

The paper addresses the fragmentation of social-media user typologies by proposing a reciprocity-based two-dimensional space defined by $r_\mathrm{in}$ and $r_\mathrm{out}$, derived from ego-network edges and degrees. Using data from 48,830 Twitter users, it shows that traditional categories like influencers and lurkers arise as regions within a continuous space rather than discrete classes, with four corner archetypes and a large intermediate region. Behavioral analyses reveal smooth gradients across reciprocity space and identify a mid-range zone where content virality peaks, challenging one-dimensional classifications. The framework offers interpretable metrics for influence and platform design, presenting a unified model of how users navigate Twitter's dual roles as information and friendship networks.

Abstract

Social media users exhibit diverse behavioral patterns as platforms function simultaneously as information and friendship networks. We introduce a reciprocity-based framework mapping users onto two-dimensional space defined by bidirectional connection ratios. Analyzing 48,830 Twitter users and 149 million connections, we demonstrate that fragmented user types from prior studies (influencers, lurkers, brokers, and follow-back accounts) emerge naturally as regions within continuous behavioral space rather than discrete categories. User properties vary smoothly across the reciprocity dimensions, revealing clear behavioral gradients. This framework provides the first unified model encompassing the full spectrum of social media behaviors and offers interpretable metrics for influence measurement and platform design.

Figures (5)

• Figure 1: User density distribution and archetype classification in reciprocity space. The two-dimensional reciprocity space shows the population distribution of 48,830 users, with $r_\mathrm{in}$ (bidirectional edges to in-degree) on the x-axis and $r_\mathrm{out}$ (bidirectional edges to out-degree) on the y-axis. Colors indicate user density, with four primary archetypes identified in the corner regions: Feeding (top left), Accumulating (bottom right), Flowing (bottom left), and Circulating (top right).
• Figure 2: (For review: To be combined with \ref{['fig:behavioral-characteristics-b', 'fig:behavioral-characteristics-c']} in final version.) Behavioral signatures and network patterns of user archetypes. Letter-value plots show distributions of user properties across five categories: Flowing (Flw), Accumulating (Acc), Feeding (Fed), Circulating (Cir), and Other (Oth). Statistical significance across all behavioral metrics was confirmed through Kruskal--Wallis tests (all $p < 4 \times 10^{-124}$). (a--d) Tweet composition patterns showing percentage of original posts, retweets, replies, and quote tweets. (e--h) Activity and temporal metrics including total posts, daily posting frequency, total likes made, and account creation dates. (i--j) Network structure showing followee and follower counts. (k--n) Content engagement metrics measuring mean retweeted and liked for all posts and original posts specifically.
• Figure 3: (For review: To be combined with \ref{['fig:behavioral-characteristics-a', 'fig:behavioral-characteristics-b']} in final version.) Behavioral signatures and network patterns of user archetypes. (a--b) Following relationship patterns showing distribution of incoming and outgoing connections by archetype, revealing intergroup interaction preferences.
• Figure 4: Continuous behavioral gradients across reciprocity space. Heatmaps showing median user property values across a $10 \times 10$ grid of the two-dimensional reciprocity space ($r_\mathrm{in}$, $r_\mathrm{out}$), with contour lines indicating smooth transitions between behavioral regions. (Caption continues on the following page.)
• Figure 5: (Continued) The analysis reveals that user behavior exists along continuous quantitative dimensions rather than discrete categories. (a--d) Tweet composition patterns showing gradual transitions in content creation behavior. (e--h) Activity and temporal patterns demonstrating smooth gradients in user engagement and platform adoption. (i--j) Network structure metrics revealing continuous variation in social connectivity patterns. (k--l) Overall content engagement showing how different reciprocity positions influence content reach. (m--n) Original content engagement metrics, with peak values occurring in an intermediate high engagement zone (approximately $r_\mathrm{in}$ 0.5--0.7, $r_\mathrm{out}$ 0.6--0.9) where balanced reciprocal connections optimize content virality. Each cell represents the median value for users within that reciprocity range.