Give the Long-tail More SPACE: Promoting Provider Fairness in Next POI Recommendation

2026-08-08

AI Summary

This work addresses the prevalent bias in existing next-point-of-interest (POI) recommendation models toward popular locations, which undermines exposure fairness for long-tail POIs and fails to account for user feasibility constraints and POI supply limitations. To mitigate this, the authors propose SPACE—a model-agnostic fairness-aware framework that enhances long-tail POI visibility by generating synthetic users who satisfy both feasibility and supply constraints. SPACE integrates community inference, imbalanced optimal transport allocation, and a constraint-guided latent diffusion mechanism, enabling seamless integration into mainstream recommendation architectures. Extensive experiments on three real-world datasets demonstrate that SPACE significantly improves provider-side fairness while maintaining or even enhancing recommendation accuracy.

Abstract

Next point-of-interest (POI) recommendation predicts users' future destinations from historical mobility sequences and has become a key component of location-based services. However, mainstream models often concentrate exposure on a small set of popular POIs, leaving long-tail merchants systematically under-exposed. While provider fairness has recently attracted increasing attention, directly applying existing provider-fairness techniques to POI recommendation is problematic: (i) users face execution constraints; and (ii) POIs face resource supply constraints. To address this, we propose SPACE (Supply- and Physics-Aware Conditional Embedding generation), a model-agnostic framework that improves long-tail POI exposure via virtual user generation under explicit feasibility and supply control. SPACE consists of three stages: (1) community inference to capture heterogeneous user execution constraints; (2) unbalanced optimal-transport allocation to decide how many virtual users each tail POI should receive from which communities under POI-specific supply budgets; and (3) constraint-guided latent diffusion to generate POI-conditional, community-consistent virtual user embeddings. The generated user-POI pairs can be seamlessly used to train existing recommenders without modifying their architectures. Extensive experiments on three real-world datasets demonstrate that SPACE substantially improves provider fairness while maintaining and often improving recommendation accuracy across multiple backbone models. Our code is publicly available at https://github.com/Anniran1/SPACE-main.

Problem

• provider fairness
• long-tail POIs
• next POI recommendation
• exposure bias
• resource constraints

Innovation

• provider fairness
• long-tail POI recommendation
• virtual user generation
• optimal transport
• constraint-aware diffusion