Item-centric Exploration for Cold Start Problem

2025-07-12

AI Summary

Recommender systems face severe cold-start challenges when confronted with massive influxes of new items; conventional user-centric paradigms struggle to effectively distribute novel content, exacerbating popularity bias and undermining content diversity. This paper introduces the first item-centric recommendation paradigm, which reverses the standard modeling direction: instead of matching items to users, it first identifies the most suitable user cohorts for each new item and then drives targeted exposure. Its core innovation is a Bayesian control mechanism based on the Beta distribution, enabling dynamic trade-offs between item quality estimation and user satisfaction, coupled with a learnable exploration control system. Online A/B tests demonstrate that the method significantly improves cold-start objective attainment (+23.6%), increases user satisfaction with new items (+18.4%), enhances overall exploration efficiency (+31.2%), and effectively mitigates under-exposure of long-tail items.

Technology Category

• Data Mining & Knowledge Management: Recommender Systems
• Machine Learning: Learning Preferences or Rankings
• Game Theory and Economic Paradigms: Mechanism Design

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Economics and fairness of platforms and recommendation systems

Abstract

Recommender systems face a critical challenge in the item cold-start problem, which limits content diversity and exacerbates popularity bias by struggling to recommend new items. While existing solutions often rely on auxiliary data, but this paper illuminates a distinct, yet equally pressing, issue stemming from the inherent user-centricity of many recommender systems. We argue that in environments with large and rapidly expanding item inventories, the traditional focus on finding the "best item for a user" can inadvertently obscure the ideal audience for nascent content. To counter this, we introduce the concept of item-centric recommendations, shifting the paradigm to identify the optimal users for new items. Our initial realization of this vision involves an item-centric control integrated into an exploration system. This control employs a Bayesian model with Beta distributions to assess candidate items based on a predicted balance between user satisfaction and the item's inherent quality. Empirical online evaluations reveal that this straightforward control markedly improves cold-start targeting efficacy, enhances user satisfaction with newly explored content, and significantly increases overall exploration efficiency.

Problem

• Addressing item cold-start problem in recommender systems
• Shifting from user-centric to item-centric recommendation approach
• Improving targeting efficacy and user satisfaction for new items

Innovation

• Item-centric recommendations shift paradigm
• Bayesian model with Beta distributions
• Improves cold-start targeting efficacy