Abstract
Global games form a subclass of games with incomplete information where a set of agents decide actions against a regime with an underlying fundamental $\theta$ representing its power. Each agent has access to an independent noisy observation of $\theta$. In order to capture the behavior of agents in a social network of information exchange we assume that agents share their observation in a noisy environment prior to making their decision. We show that global games with noisy sharing of information do not admit an intuitive type of threshold policy which only depends on agents' belief about the underlying $\theta$. This is in contrast to the existing results on the threshold policy for the conventional set-up of global games. Motivated by this result, we investigate the existence of equilibrium strategies in a more general collection of threshold-type policies and show that such equilibrium strategies exist and are unique if the sharing of information happens over a sufficiently noisy environment.
Introduction
Global games provide a robust framework for modeling coordination problems under uncertainty, such as bank runs, currency crises, and political revolutions. In these games, agents must decide whether to take a risky action against a "regime" based on their private, noisy observations of the regime's strength, denoted by $\theta$. Success depends on whether the aggregate action of the agents exceeds a threshold determined by the true value of $\theta$.
Traditional global game models assume that agents' observations are independent. However, modern social interactions, particularly through digital networks, involve significant information exchange. This paper extends the standard model by introducing a "noisy sharing" phase, where agents communicate their private signals over imperfect channels before making their final decision.
Information Structure
The model assumes a set of $n$ agents. The process begins with each agent $i$ receiving a private signal $x_i$ regarding the fundamental $\theta$, where $x_i = \theta + \epsilon_i$ and $\epsilon_i \sim \mathcal{N}(0, \sigma^2)$. Following this, agents share their signals. Due to environmental noise or communication constraints, the shared signals are further corrupted. Agent $i$ receives a signal $y_{ji}$ from agent $j$, modeled as $y_{ji} = x_j + \eta_{ji}$, where $\eta_{ji} \sim \mathcal{N}(0, \tau^2)$.
As shown in Figure 1, this creates a two-tier noise structure. Each agent $i$ ultimately possesses a vector of observations $\mathbf{y}_i = [y_{1i}, y_{2i}, \dots, y_{ni}]^T$, where $y_{ii} = x_i$. The challenge lies in determining how agents should map this multi-dimensional information into a binary decision (to attack or not).
The Failure of Intuitive Strategies
In standard global games, a common "intuitive" strategy is for an agent to calculate their posterior expectation of the fundamental, $E[\theta | \mathbf{y}_i]$, and act if this value falls below a certain threshold. One of the primary contributions of this work is Theorem 1, which proves that such a strategy does not constitute a Bayesian Nash Equilibrium (BNE) in the presence of noisy sharing. The interdependence introduced by the sharing phase breaks the simple scalar threshold logic that characterizes independent-signal global games.
Existence and Uniqueness of Equilibrium
Given the failure of simple strategies, the authors investigate a broader class of threshold-type policies. They define a threshold function $h(\mathbf{y}_i)$ and seek a symmetric equilibrium where all agents adopt the same rule.
By formulating the equilibrium condition as a functional fixed-point equation, the authors apply the Banach Fixed Point Theorem. They establish (Theorem 3) that a unique, symmetric threshold policy equilibrium exists provided the communication environment is "sufficiently noisy." In this context, noise acts as a regularizing force that ensures the mapping between strategies is a contraction, leading to a stable and predictable outcome.
Conclusion and Future Work
This research demonstrates that while information sharing complicates the strategic landscape of global games, stability can still be achieved under specific noise conditions. The finding that higher noise levels can actually facilitate the convergence to a unique equilibrium is particularly striking, suggesting that noise can act as a stabilizing factor in social coordination.
Future work may extend these findings to non-Gaussian noise distributions or more complex network topologies where information sharing is not all-to-all. Additionally, exploring dynamic, multi-round versions of these games could provide deeper insights into how social coordination evolves over time in noisy environments.
Reference: Mahdavifar, H., Beirami, A., Touri, B., & Shamma, J. S. (2017). Global Games with Noisy Information Sharing. IEEE Transactions on Signal and Information Processing over Networks.
Summary updated