Avoiding Over-Personalization with Rule-Guided Knowledge Graph Adaptation for LLM Recommendations

2025-09-08

AI Summary

To address filter bubbles and information narrowing caused by excessive personalization in large language model (LLM)-based recommender systems, this paper proposes a lightweight neuro-symbolic framework. During inference, it dynamically reconstructs user-side knowledge graphs and applies rule-guided graph adaptation—including triplet reweighting, hard inversion, and bias removal—to mitigate feature co-occurrence bias. Coupled with structured prompt engineering and client-side customized optimization, the approach jointly enhances recommendation diversity and topic relevance. Its core innovation lies in enabling controllable adjustment of personalization intensity at inference time—without any model fine-tuning—through symbolic graph operations and prompt guidance alone. Evaluated on a recipe recommendation benchmark, the method achieves a +23.6% improvement in novelty while preserving recommendation accuracy, outperforming both global adaptation and naive prompting baselines.

Technology Category

• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Machine Learning: Learning Preferences or Rankings
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

We present a lightweight neuro-symbolic framework to mitigate over-personalization in LLM-based recommender systems by adapting user-side Knowledge Graphs (KGs) at inference time. Instead of retraining models or relying on opaque heuristics, our method restructures a user's Personalized Knowledge Graph (PKG) to suppress feature co-occurrence patterns that reinforce Personalized Information Environments (PIEs), i.e., algorithmically induced filter bubbles that constrain content diversity. These adapted PKGs are used to construct structured prompts that steer the language model toward more diverse, Out-PIE recommendations while preserving topical relevance. We introduce a family of symbolic adaptation strategies, including soft reweighting, hard inversion, and targeted removal of biased triples, and a client-side learning algorithm that optimizes their application per user. Experiments on a recipe recommendation benchmark show that personalized PKG adaptations significantly increase content novelty while maintaining recommendation quality, outperforming global adaptation and naive prompt-based methods.

Problem

• Mitigate over-personalization in LLM recommender systems
• Suppress feature co-occurrence patterns reinforcing filter bubbles
• Increase content novelty while maintaining recommendation quality

Innovation

• Lightweight neuro-symbolic framework adapts user knowledge graphs
• Symbolic strategies reweight invert remove biased graph triples
• Structured prompts steer LLM toward diverse relevant recommendations