When Knowing Early Matters: Gossip, Percolation and Nash Equilibria

Continually arriving information is communicated through a network of $n$ agents, with the value of information to the $j$‘th recipient being a decreasing function of $j/n$, and communication costs paid by recipient. Regardless of details of network and communication costs, the social optimum policy is to communicate arbitrarily slowly. But selfish agent behavior leads to Nash equilibria which (in the $n \to \infty$ limit) may be efficient (Nash payoff $=$ social optimum payoff) or wasteful ($0 < $ Nash payoff $<$ social optimum payoff) or totally wasteful (Nash payoff $=0$). We study the cases of the complete network (constant communication costs between all agents), the grid with only nearest-neighbor communication, and the grid with communication cost a function of distance. The main technical tool is analysis of the associated first passage percolation process or SI epidemic (representing spread of one item of information) and in particular its “window width”, the time interval during which most agents learn the item. Many arguments are just outlined, not intended as complete rigorous proofs.

💡 Research Summary

The paper studies a dynamic information‑sharing environment in which an infinite stream of items arrives and spreads through a network of n agents. Each item’s value to the j‑th recipient is a decreasing function of the rank j/n, while the recipient bears the communication cost. The authors first characterize the social optimum: regardless of network topology or cost structure, the welfare‑maximizing policy is to transmit information arbitrarily slowly, because delaying transmission preserves higher per‑recipient value and reduces total cost.

Self‑interested agents, however, choose when to receive each item to maximize their own net payoff. This strategic interaction is modeled as a non‑cooperative game, and Nash equilibria are examined in the limit n → ∞. Three equilibrium regimes emerge:

1. Efficient Nash – the equilibrium payoff equals the social optimum; agents’ selfish timing does not degrade overall efficiency.
2. Partially wasteful Nash – the equilibrium payoff is positive but strictly below the optimum; some efficiency is lost.
3. Totally wasteful Nash – the equilibrium payoff collapses to zero; the system becomes completely inefficient.

The analysis hinges on the first‑passage percolation (or SI epidemic) process that describes the spread of a single information item. A central concept is the “window width”: the time interval during which the majority of agents acquire the item. A narrow window corresponds to rapid, burst‑like spread, which forces agents to act quickly and often leads to wasteful equilibria. A wide window indicates a gradual diffusion, allowing agents to delay receipt without sacrificing value, thereby supporting efficient equilibria.

The authors apply this framework to three canonical network settings:

• Complete graph – every pair of agents can communicate at a constant cost. The percolation front advances almost instantaneously, yielding a window width of order 1/n. Consequently, agents experience strong incentives to receive items immediately, producing either partially wasteful or totally wasteful equilibria depending on the magnitude of the uniform cost.

• Nearest‑neighbor 2‑D grid – agents communicate only with the four adjacent nodes. The spread follows a diffusive front with a window width of order 1, which is large enough for agents to wait for higher‑value reception. In this case an efficient Nash equilibrium exists; the selfish outcome coincides with the social optimum.

• Distance‑dependent cost grid – communication cost grows as a function f(d) of Euclidean distance d. The authors identify a phase transition: if f(d) grows faster than linearly, the percolation front is severely slowed, the window narrows, and totally wasteful equilibria dominate. If f(d) grows sub‑linearly (e.g., logarithmically), the front remains relatively fast, the window stays wide, and efficient equilibria persist. Intermediate growth rates generate partially wasteful outcomes.

To obtain quantitative bounds on the window width, the paper blends large‑deviation techniques, variational principles, and results from first‑passage percolation theory. While many proofs are sketched rather than fully formalized, the methodological core is clear: the interaction between transmission speed and cost structure determines the shape of the percolation front, which in turn dictates the equilibrium class.

The findings have practical implications for the design of communication protocols, social‑media platforms, and distributed systems where timely information is valuable but costly to disseminate. By adjusting cost functions—e.g., imposing distance‑based fees or congestion‑based pricing—system designers can manipulate the percolation window and steer selfish behavior toward socially efficient outcomes. Conversely, ignoring these incentives may lead to severe inefficiencies, especially in highly connected networks where rapid spread is cheap.

The paper concludes by suggesting extensions: multi‑item simultaneous diffusion, heterogeneous value functions, dynamic or stochastic cost models, and richer network topologies. Such directions would deepen the connection between game theory, epidemic modeling, and network science, offering a fertile ground for future research.