Fair Influence Maximization in Hypergraphs

Abstract

Influence maximization (IM) finds applications in viral marketing, public health campaigns, and social influence. However, in networks with strong community structure, standard IM strategies can create severe disparities, systematically under-reaching certain groups. While fairness in IM has been studied for pairwise networks, real-world systems often exhibit higher-order interactions (e.g., group chats, co-authorship, meetings) that are naturally modeled as hypergraphs, yet fairness in this setting remains largely unexplored.
We introduce FIMH, an algorithm for fair influence maximization on hypergraphs. Operating under the Susceptible–Infected Contact Process (SICP) model, FIMH jointly optimizes total influence and fairness across communities using a structural influence estimation and a parameter-free utopia-distance selection criterion. Experiments on seven real-world hypergraph datasets demonstrate that FIMH has competitive influence performance to state-of-the-art methods while reducing inter-community disparity by 31% on average and up to 52%. Our results establish that fairness and influence are not competing objectives in hypergraph diffusion such that balanced information spread can be achieved without sacrificing reach.