A Survey of Graph Neural Networks for Social Recommender Systems

TL;DR

The paper addresses the challenge of designing effective social recommender systems by exploiting both user-item interactions and user-user social relations through graph neural networks. It introduces a novel taxonomy for inputs and architectures, classifying methods into 8 encoder families, 2 decoders, and 12 loss-function categories, and catalogs 17 benchmark datasets and common evaluation metrics. The survey synthesizes primary losses (MSE, BPR, CE, hinge) and auxiliary signals (SSL, LP, KD, etc.), analyzes encoder/decoder trade-offs, and discusses time complexity across methods. It also outlines future directions, including graph augmentation, trustworthy GNNs, heterogeneity, and scalability, to guide researchers toward robust and scalable SocialRS solutions with real-world impact.

Abstract

Social recommender systems (SocialRS) simultaneously leverage the user-to-item interactions as well as the user-to-user social relations for the task of generating item recommendations to users. Additionally exploiting social relations is clearly effective in understanding users' tastes due to the effects of homophily and social influence. For this reason, SocialRS has increasingly attracted attention. In particular, with the advance of graph neural networks (GNN), many GNN-based SocialRS methods have been developed recently. Therefore, we conduct a comprehensive and systematic review of the literature on GNN-based SocialRS. In this survey, we first identify 84 papers on GNN-based SocialRS after annotating 2151 papers by following the PRISMA framework (preferred reporting items for systematic reviews and meta-analyses). Then, we comprehensively review them in terms of their inputs and architectures to propose a novel taxonomy: (1) input taxonomy includes 5 groups of input type notations and 7 groups of input representation notations; (2) architecture taxonomy includes 8 groups of GNN encoder notations, 2 groups of decoder notations, and 12 groups of loss function notations. We classify the GNN-based SocialRS methods into several categories as per the taxonomy and describe their details. Furthermore, we summarize benchmark datasets and metrics widely used to evaluate the GNN-based SocialRS methods. Finally, we conclude this survey by presenting some future research directions. GitHub repository with the curated list of papers are available at https://github.com/claws-lab/awesome-GNN-social-recsys.

Figures (9)

• Figure 1: The number of papers related to GNN-based SocialRS per year. $^\star$ For 2022, we count the number of relevant papers published until October.
• Figure 2: A timeline of GNN-based SocialRS methods. We categorize methods according to their GNN encoders: graph convolutional network (GCN), lightweight GCN (LightGCN), graph attention neural networks (GANN), heterogeneous GNN (HetGNN), graph recurrent neural networks (GRNN), hypergraph neural networks (HyperGNN), graph autoencoder (GAE), and hyperbolic GNN. It should be noted that some methods employ two or more GNN encoders in their architectures.
• Figure 3: The number of GNN-based SocialRS papers published in relevant journals and conferences. We only present statistics with respect to prominent data mining journals (including IEEE TKDE, ACM TOIS, Knowledge-Based Systems, and Information Sciences) and conferences (including WWW, ACM SIGIR, ACM KDD, ACM CIKM, ACM WSDM, IEEE ICDE, and IEEE ICDM). We believe it would help researchers in this field to identify appropriate venues where GNN-based SocialRS papers are published.
• Figure 4: Overview of input types used by GNN-based SocialRS methods.
• Figure 5: Overview of input representations used by GNN-based SocialRS methods.