An evidence-accumulating drift-diffusion model of competing information spread on networks

TL;DR

The paper develops an evidence-accumulating drift-diffusion model for competing information on networks, where each agent maintains a stance $s_i(t)\in\{-1,0,+1\}$ and a confidence $c_i(t)\in\mathbb{R}$ that evolves under social input, memory decay, and noise. By conducting extensive Monte Carlo simulations across Random-Regular, Watts-Strogatz, Barabási–Albert, and Stochastic Block Model networks, the authors show that the persistence of information sources (duration $\tau$) and the number of committed agents ($\zeta^+$, $\zeta^-$) jointly shape the emergent population-level behaviours: single sources lead to consensus, while opposing sources can yield either consensus or polarisation depending on parameter regimes, with persistence playing a stronger role than quantity. They introduce entropy-based metrics $\bar{X}$, $\bar{Y}$, and a network-disagreement measure $\bar{D}$ to classify behaviours and demonstrate robustness of findings across topologies and larger networks. The work advances understanding of how memory and gradual evidence accumulation influence information cascades and echo-chamber formation, offering insights for interventions aimed at mitigating polarisation in real-world networks.

Abstract

In this paper, we propose an agent-based model of information spread, grounded on psychological insights on the formation and spread of beliefs. In our model, we consider a network of individuals who share two opposing types of information on a specific topic (e.g., pro- vs. anti-vaccine stances), and the accumulation of evidence supporting either type of information is modelled by means of a drift-diffusion process. After formalising the model, we put forward a campaign of Monte Carlo simulations to identify population-wide behaviours emerging from agents' exposure to different sources of information, investigating the impact of the number and persistence of such sources, and the role of the network structure through which the individuals interact. We find similar emergent behaviours for all network structures considered. When there is a single type of information, the main observed emergent behaviour is consensus. When there are opposing information sources, both consensus or polarisation can result; the latter occurs if the number and persistence of the sources exceeds some threshold values. Importantly, we find the emergent behaviour is mainly influenced by how long the information sources are present for, as opposed to how many sources there are.

Figures (8)

• Figure 1: Schematic of the agents' dynamics.
• Figure 2: Temporal evolution of the fraction of agents sharing $+1$ (solid blue) and $-1$ (dashed orange) for different simulations with the same set of parameters and initial conditions.
• Figure 3: Characterisation of possible emergent behaviours of the model as a function of temporal ($\bar{X}$) and spatial ($\bar{Y}$) entropy.
• Figure 4: Results of the Monte Carlo simulations for a RR network in the single-source case ($\zeta^- = 0$). In panels (a), (b), and (c), we report the values of the temporal entropy $\bar{X}$, spatial entropy $\bar{Y}$, and average network disagreement $\bar{D}$, respectively, using a colour code, for different agent commitment durations $\tau$ and fractions of committed agents $\zeta^+$. In panels (d--f), we represent the temporal evolution of the network disagreement $D(t)$ for three pairs of meta-parameters $(\tau,\zeta^+)$ of interest, represented by the boxed areas in panel (c).
• Figure 5: Results of the Monte Carlo simulations for a RR network in the dual-source case with equal proportions of committed agents of each type ($\zeta^- = \zeta^+ = \zeta^\pm$). In panels (a), (b), and (c), we report the values of the temporal entropy $\bar{X}$, spatial entropy $\bar{Y}$, and average network disagreement $\bar{D}$, respectively, using a colour code, for different agent commitment durations $\tau$ and fractions of committed agents $\zeta^+$. In panels (d--f), we represent the temporal evolution of the network disagreement $D(t)$ for three pairs of meta-parameters $(\tau,\zeta^+)$ of interest, represented by the boxed areas in panel (c).