Overview
Research investigated a stochastic opinion dynamics model within a fully connected social network. This model involves $N$ actors interacting and expressing opinions from a set of $M$ discrete options. The study focused on the evolution of social pressure exerted on each actor for each opinion, which is conceptualized as a Markov jump process. Key outcomes include the observation of fast consensus formation, the establishment of a unique invariant measure, and the presence of metastability under conditions of high network polarization. Furthermore, the introduction of a communication bias parameter revealed a phase transition within the system dynamics.
Research Context
The model considers a social network where interactions lead to the formation and evolution of opinions. Each actor maintains an $M$-tuple, representing the social pressure they experience for each of the $M$ possible opinions. This internal pressure dictates an actor's tendency to express opinions, a tendency which is modulated by a polarization coefficient. The framework describes the dynamics of opinion expression and the consequent shifts in social pressure across the network.
Approach
The system's evolution is modeled as a Markov jump process, tracking the matrix of social pressures for all actors across all opinions. When an actor expresses a specific opinion, say opinion $o$, their individual social pressure for all opinions is reset to zero. Concurrently, for all other actors in the network, the social pressure for opinion $o$ increases by 1 unit, while the social pressure for all other opinions distinct from $o$ decreases by a factor of $1/(M-1)$. The model incorporates a polarization coefficient to modulate an actor's tendency to express opinions based on their social pressure vector. Additionally, a communication bias parameter was introduced to investigate its impact on the system's long-term behavior.
Findings
-
Fast Consensus Formation: The model demonstrates the rapid emergence of consensus among actors within the network.
-
Unique Invariant Measure: The system possesses a unique invariant measure, suggesting a stable long-term statistical behavior.
-
Metastability in Polarized Networks: In networks characterized by high polarization, the system exhibits metastability, indicating the presence of multiple quasi-stable states where the system can temporarily reside before transitioning to another state.
-
Phase Transition due to Communication Bias: The introduction of a communication bias parameter leads to a distinct phase transition:
- If the communication bias parameter is negative, all actors except for one cease expressing opinions within a finite time, almost surely.
- If the communication bias parameter is not negative (i.e., zero or positive), no actor stops expressing opinions.