Leveraging LLMs for Influence Path Planning in Proactive Recommendation

TL;DR

LLM-IPP tackles echo chambers in recommender systems by guiding user interest toward a target item via an influence path. It leverages prompt-engineered LLMs to generate coherent paths that include the target item, with dataset-specific constraints and optional CoT/ToT prompting. The authors introduce both traditional and LLM-based evaluation metrics and simulators, and demonstrate that LLM-IPP improves path coherence and user acceptability over IRS baselines on MovieLens-1M and Last.FM datasets, including case studies and human evaluation. The work establishes a new direction for proactive recommendation by integrating LLM planning capabilities and suggests future work on multi-round evaluations and memory-augmented prompts.

Abstract

Recommender systems are pivotal in Internet social platforms, yet they often cater to users' historical interests, leading to critical issues like echo chambers. To broaden user horizons, proactive recommender systems aim to guide user interest to gradually like a target item beyond historical interests through an influence path,i.e., a sequence of recommended items. As a representative, Influential Recommender System (IRS) designs a sequential model for influence path planning but faces issues of lacking target item inclusion and path coherence. To address the issues, we leverage the advanced planning capabilities of Large Language Models (LLMs) and propose an LLM-based Influence Path Planning (LLM-IPP) method. LLM-IPP generates coherent and effective influence paths by capturing user interest shifts and item characteristics. We introduce novel evaluation metrics and user simulators to benchmark LLM-IPP against traditional methods. Our experiments demonstrate that LLM-IPP significantly enhances user acceptability and path coherence, outperforming existing approaches.

Figures (1)

• Figure 1: Illustration of Passive and Proactive Recommendation.