How Agreement and Disagreement Evolve over Random Dynamic Networks

The dynamics of an agreement protocol interacting with a disagreement process over a common random network is considered. The model can represent the spreading of true and false information over a communication network, the propagation of faults in a large-scale control system, or the development of trust and mistrust in a society. At each time instance and with a given probability, a pair of network nodes are selected to interact. At random each of the nodes then updates its state towards the state of the other node (attraction), away from the other node (repulsion), or sticks to its current state (neglect). Agreement convergence and disagreement divergence results are obtained for various strengths of the updates for both symmetric and asymmetric update rules. Impossibility theorems show that a specific level of attraction is required for almost sure asymptotic agreement and a specific level of repulsion is required for almost sure asymptotic disagreement. A series of sufficient and/or necessary conditions are then established for agreement convergence or disagreement divergence. In particular, under symmetric updates, a critical convergence measure in the attraction and repulsion update strength is found, in the sense that the asymptotic property of the network state evolution transits from agreement convergence to disagreement divergence when this measure goes from negative to positive. The result can be interpreted as a tight bound on how much bad action needs to be injected in a dynamic network in order to consistently steer its overall behavior away from consensus.

💡 Research Summary

The paper introduces a stochastic interaction model for a network of agents in which, at each discrete time step, a randomly selected pair of nodes updates its state according to one of three possible rules: attraction (moving toward the partner’s state), repulsion (moving away from the partner’s state), or neglect (remaining unchanged). The probability of selecting any given pair is uniform, and the choice of rule for each node is independent and governed by prescribed probabilities. The authors study both symmetric updates—where both nodes use the same attraction and repulsion strengths—and asymmetric updates—where each node may have its own parameters or only one node updates.

Using Markov chain theory and stochastic matrix products, the authors derive conditions under which the entire network converges almost surely to consensus (agreement) or diverges almost surely to disagreement. A key contribution is an “impossibility theorem” that establishes a minimum level of attraction required for almost‑sure consensus and a minimum level of repulsion required for almost‑sure divergence. In the symmetric case, they define a critical convergence measure γ = α – β, where α is the attraction strength and β the repulsion strength. When γ < 0 the system converges to agreement; when γ > 0 it diverges to disagreement; γ = 0 marks a delicate boundary where the outcome depends on initial conditions. This measure is linked to the second smallest eigenvalue λ₂ of the graph Laplacian, showing that more strongly connected graphs tolerate larger repulsion before tipping into disagreement.

For asymmetric updates, the analysis reveals that while the average behavior follows the same γ‑criterion, individual nodes may converge at different rates. Nodes with higher update frequencies (the “core” nodes) can drive the network toward consensus even when many peripheral nodes are repelling, but persistent repulsion by a minority can create localized clusters of disagreement. The paper provides both sufficient and necessary conditions for convergence/divergence, expressed in terms of spectral radii of the expected transition matrices.

Extensive simulations on complete graphs, Erdős–Rényi random graphs, and Barabási–Albert scale‑free networks validate the theoretical predictions. The simulations demonstrate a sharp phase transition at γ = 0, with the transition point shifting according to λ₂: dense, well‑connected graphs require stronger repulsion to cause divergence, whereas sparse, heterogeneous graphs diverge with relatively weak repulsion. Asymmetric scenarios show that increasing the proportion of high‑frequency “influencer” nodes reduces the time to consensus, while a small fraction of stubborn repelling nodes can sustain disagreement pockets.

The results have practical implications for designing resilient communication or control systems, for understanding the spread of misinformation versus factual information, and for modeling trust dynamics in societies. By quantifying the exact balance of attractive and repulsive forces needed to steer a network toward or away from consensus, the work offers a rigorous tool for policymakers and engineers to assess how much “bad action” must be injected to reliably disrupt agreement, or conversely, how much reinforcement is necessary to guarantee convergence despite adversarial influences. Future extensions suggested include time‑varying topologies, multi‑state opinion models, and external control inputs such as targeted advertising or cyber‑attacks.