Maximizing Welfare in Social Networks under a Utility Driven Influence Diffusion Model

TL;DR

This paper introduces a new influence diffusion model that incorporates user utility and complements influence maximization with social welfare optimization, addressing economic and multi-item considerations.

Contribution

It proposes the UIC model combining utility-driven adoption with influence propagation, and develops a scalable greedy algorithm with strong approximation guarantees.

Findings

The greedy algorithm achieves near-optimal social welfare approximation.

BundleGRD outperforms baseline methods in experiments.

The model effectively integrates economic utility into influence maximization.

Abstract

Motivated by applications such as viral marketing, the problem of influence maximization (IM) has been extensively studied in the literature. The goal is to select a small number of users to adopt an item such that it results in a large cascade of adoptions by others. Existing works have three key limitations. (1) They do not account for economic considerations of a user in buying/adopting items. (2) Most studies on multiple items focus on competition, with complementary items receiving limited attention. (3) For the network owner, maximizing social welfare is important to ensure customer loyalty, which is not addressed in prior work in the IM literature. In this paper, we address all three limitations and propose a novel model called UIC that combines utility-driven item adoption with influence propagation over networks. Focusing on the mutually complementary setting, we formulate the problem of social welfare maximization in this novel setting. We show that while the objective function is neither submodular nor supermodular, surprisingly a simple greedy allocation algorithm achieves a factor of $(1-1/e-\epsilon)$ of the optimum expected social welfare. We develop \textsf{bundleGRD}, a scalable version of this approximation algorithm, and demonstrate, with comprehensive experiments on real and synthetic datasets, that it significantly outperforms all baselines.