Finding safe strategies for competitive diffusion on trees

TL;DR

This paper investigates safe strategies in a competitive influence diffusion game on trees, providing bounds, algorithms, and empirical tests to maximize influence spread under uncertainty.

Contribution

It introduces a method to compute safe strategies on trees, with tight bounds and an algorithm that performs near-optimally in practice.

Findings

Tight bounds on minimal expected gain for spiders and complete trees.

An algorithm for safe strategies applicable to any tree.

Strategies found are close to optimal in experiments.

Abstract

We study the two-player safe game of Competitive Diffusion, a game-theoretic model for the diffusion of technologies or influence through a social network. In game theory, safe strategies are mixed strategies with a minimal expected gain against unknown strategies of the opponents. Safe strategies for competitive diffusion lead to maximum spread of influence in the presence of uncertainty about the other players. We study the safe game on two specific classes of trees, spiders and complete trees, and give tight bounds on the minimal expected gain. We then use these results to give an algorithm which suggests a safe strategy for a player on any tree. We test this algorithm on randomly generated trees, and show that it finds strategies that are close to optimal.