arXiv cs.CY· Amir Ghasemian, Homa Hosseinmardi, Upasana Dutta, Duncan J. Watts·· 23 小时前AI 评分40
LLM 辅助重排序如何在推荐系统中落实细微目标
LLM-Assisted Reranking to Operationalize Nuanced Objectives in Recommender Systems
AI 导读
研究用真实 YouTube 轨迹,以零样本提示对三个模型及一个嵌入基线重排侧边栏候选,评估相关性、意识形态对齐与阴谋/极端内容暴露。纯个性化指令下重排序提升相关性但收窄意识形态暴露,三个模型中有两个会向有此类历史的用户推送问题内容,第三个模型因更明确拒绝党派请求体现模型特定护栏。轻量提示级约束在三个模型上均大幅降低问题内容暴露,且正则化重排器相关性仍高于平台基线。
正文 · 原文
Abstract:Recommender systems shape what information we see, raising concerns about filter bubbles, radicalization, polarization, and social inequality. Large language models (LLMs) enable more powerful personalization that may intensify these dynamics, yet most recommenders are tuned for engagement or narrow accuracy metrics, with little attention to broader social effects. We investigate whether LLM-assisted reranking, while improving personalization, inadvertently amplifies exposure to ideologically congenial, extreme, or conspiratorial political content, a risk theorized but not yet empirically characterized. Using real YouTube trajectories, we rerank the platform's sidebar candidates by zero-shot prompting across three models, alongside an embedding baseline, and evaluate relevance, ideological alignment and diversity, and exposure to conspiratorial or extremist content. We find that (i) under a pure personalization instruction, reranking improves relevance while narrowing ideological exposure and, on two of three models, promoting problematic content for users with such histories; the third model, which declines most explicitly partisan requests, points to model-specific guardrails; (ii) lightweight prompt-level constraints substantially reduce problematic exposure on all three models, while the regularized rerankers still rank above the platform baseline on relevance; and (iii) synthetic experiments varying topical and partisan cues independently show no fixed preference between them: topic dominates on abortion and immigration, while partisanship takes over on elections, where the topical signal is weaker. Our results highlight the power of LLMs to operationalize contextual nuance in high-stakes political recommendation, and underscore the need to evaluate LLM-assisted personalization beyond accuracy and to treat prompt design as a value-laden choice rather than a neutral default.
| Comments: | 42 pages total; 11 pages, 5 figures, 0 tables (main text); 31 pages, 19 figures, 24 tables (appendix) |
| Subjects: | Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Information Retrieval (cs.IR) |
| Cite as: | arXiv:2606.02883 [cs.HC] |
| (or arXiv:2606.02883v2 [cs.HC] for this version) | |
| https://doi.org/10.48550/arXiv.2606.02883 arXiv-issued DOI via DataCite |
Submission history
From: Amir Ghasemian [view email]
[v1]
Mon, 1 Jun 2026 20:54:14 UTC (6,522 KB)
[v2]
Wed, 18 Nov 2026 02:24:43 UTC (6,604 KB)
来源:arXiv cs.CY · arxiv.org