Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A dynamical model of platform choice and online segregation

Sven Banisch, Dennis Jacob, Tom Willaert, Eckehard Olbrich

TL;DR

This study develops a dynamic model of platform selection, extending Social Feedback Theory by incorporating multi-agent reinforcement learning to capture how user decisions are shaped by past rewards across different platforms.

Abstract

In order to truly understand how social media might shape online discourses or contribute to societal polarization, we need refined models of platform choice, that is: models that help us understand why users prefer one social media platform over another. This study develops a dynamic model of platform selection, extending Social Feedback Theory by incorporating multi-agent reinforcement learning to capture how user decisions are shaped by past rewards across different platforms. A key parameter ($μ$) in the model governs users' tendencies to either seek approval from like-minded peers or engage with opposing views. Our findings reveal that online environments can evolve into suboptimal states characterized by polarized, strongly opinionated echo chambers, even when users prefer diverse perspectives. Interestingly, this polarizing state coexists with another equilibrium, where users gravitate toward a single dominant platform, marginalizing other platforms into extremity. Using agent-based simulations and dynamical systems analysis, our model underscores the complex interplay of user preferences and platform dynamics, offering insights into how digital spaces might be better managed to foster diverse discourse.

A dynamical model of platform choice and online segregation

TL;DR

This study develops a dynamic model of platform selection, extending Social Feedback Theory by incorporating multi-agent reinforcement learning to capture how user decisions are shaped by past rewards across different platforms.

Abstract

In order to truly understand how social media might shape online discourses or contribute to societal polarization, we need refined models of platform choice, that is: models that help us understand why users prefer one social media platform over another. This study develops a dynamic model of platform selection, extending Social Feedback Theory by incorporating multi-agent reinforcement learning to capture how user decisions are shaped by past rewards across different platforms. A key parameter () in the model governs users' tendencies to either seek approval from like-minded peers or engage with opposing views. Our findings reveal that online environments can evolve into suboptimal states characterized by polarized, strongly opinionated echo chambers, even when users prefer diverse perspectives. Interestingly, this polarizing state coexists with another equilibrium, where users gravitate toward a single dominant platform, marginalizing other platforms into extremity. Using agent-based simulations and dynamical systems analysis, our model underscores the complex interplay of user preferences and platform dynamics, offering insights into how digital spaces might be better managed to foster diverse discourse.

Paper Structure

This paper contains 30 sections, 1 theorem, 26 equations, 9 figures.

Table of Contents

  1. Introduction
  2. Agent-Based Model
  3. Platform choice.
  4. Platform activity and opinion.
  5. Rewards.
  6. Agent update.
  7. Platform rewards.
  8. Simulations
  9. Model phenomenology
  10. Simulation setting.
  11. Preference for positive feedback only ($\mu = 0$).
  12. Striving for diversity only ($\mu = 1$).
  13. Striving for diversity and homophily ($\mu = 0.7$).
  14. Complex transient dynamics.
  15. Platform rewards
  16. ...and 15 more sections

Key Result

Lemma 1

The bifurcation diagrams for $\Delta Q^+$ and $\Delta Q^-$ as a function of $\mu$ are both symmetric across the value $0$. Furthermore, the bifurcation diagrams will be equivalent.

Figures (9)

  • Figure 1: Platform choice models address the co-evolutionary process of users’ preferences and decisions to engage with different platforms, and the resulting platform ecology in which these decisions take place.
  • Figure 2: Example run for $\mu = 0.0$ which shows platform polarization due to homophily, with opinions clustering on separate platforms. The evolution of the normalized platform activity for the five platforms ($A_k(t)/N$) is shown in the top panel, and the platform opinion ($O_k(t)$) is shown below. The proportion of silent agents is also shown in the activity plot by the black curve (very close to zero). Other parameters: $N = 500, M = 5, \alpha = 0.1, \beta = 8$.
  • Figure 3: Example run for $\mu = 1.0$, illustrating balanced platform activity with diverse, non-opinionated platforms when agents prioritize diversity. In this parameter setting, all platforms approach an average opinion $O_k \approx 0$ and are hence not opinionated. Other parameters: $N = 500, M = 5, \alpha = 0.1, \beta = 8$.
  • Figure 4: First run for $\mu = 0.7$, demonstrating polarization after an initial phase of platform competition (referred to as Run A). Other parameters: $N = 500, M = 5, \alpha = 0.1, \beta = 8$.
  • Figure 5: Second run for $\mu = 0.7$, showing the emergence of a single diverse platform (mega-platform) while others become marginalized (referred to as Run B). Other parameters: $N = 500, M = 5, \alpha = 0.1, \beta = 8$.
  • ...and 4 more figures

Theorems & Definitions (3)

  • Claim 1: (Fixed points are symmetric across line $y = x$)
  • Claim 2: (Fixed points are symmetric across line $y = -x$)
  • Lemma 1: (Bifurcation plots for $\Delta Q^+$ and $\Delta Q^-$ are equivalent)