The Friendship Paradox and Social Network Participation

TL;DR

This paper explores how social comparison driven by the friendship paradox influences content sharing behaviors in online networks, showing that responses can either decrease or sustain overall sharing depending on their nature.

Contribution

It introduces a simulation framework to analyze how social comparison impacts sharing dynamics, highlighting the effects of different behavioral response types.

Findings

Monotonic responses lead to declining sharing over time.

Convex responses can maintain or increase sharing levels.

Simulations are based on synthetic network models, not real-world data.

Abstract

The friendship paradox implies that a person will, on average, have fewer friends than their friends do. Prior work has shown how the friendship paradox can lead to perception biases regarding behaviors that correlate with the number of friends: for example, people tend to perceive their friends as being more socially engaged than they are. Here, we investigate the consequences of this type of social comparison in the conceptual setting of content creation ("sharing") in an online social network. Suppose people compare the amount of feedback that their content receives to the amount of feedback that their friends' content receives, and suppose they modify their sharing behavior as a result of that comparison. How does that impact overall sharing on the social network over time? We run simulations over model-generated synthetic networks, assuming initially uniform sharing and feedback rates. Thus, people's initial modifications of their sharing behavior in response to social comparisons are entirely driven by the friendship paradox. These modifications induce inhomogeneities in sharing rates that can further alter perception biases. If people's responses to social comparisons are monotonic (i.e., the larger the disparity, the larger the modification in sharing behavior), our simulations suggest that overall sharing in the network gradually declines. Meanwhile, convex responses can sustain or grow overall sharing in the network. We focus entirely on synthetic graphs in the present work and have not yet extended our simulations to real-world network topologies. Nevertheless, we do discuss practical implications, such as how interventions can be tailored to sustain long-term sharing, even in the presence of adverse social-comparison effects.