AuditAgent: LLM Agent for Risks Auditing in Recommender Systems
Du Su, Zhenxing Chen, Shilong Zhao, Yuanhao Liu, Fei Sun, Qi Cao, Huawei Shen
Abstract
Auditing recommendation systems has attracted growing attention due to increasing concerns over filter bubbles, unfairness, and data misuse. A common approach is sock-puppet auditing, where autonomous agents interact with platforms to reveal risks. However, existing approaches rely on hard-coded agents, lacking adaptability to dynamic GUI layouts and generating behaviors far from those of real users, limiting the comprehensiveness and representativeness of assessment. To address these issues, we introduce AuditAgent, an LLM-powered GUI-agent framework for risk auditing. AuditAgent simulates realistic user preferences and performs adaptive, human-like interactions on recommendation platforms. This design enables more thorough and faithful auditing, providing comprehensive assessments across multiple risk dimensions, including filter bubbles, unfairness, and data misuse.
BibTeX
@inproceedings{aaai2026_auditagentllmage,
title = {AuditAgent: LLM Agent for Risks Auditing in Recommender Systems},
author = {Du Su and Zhenxing Chen and Shilong Zhao and Yuanhao Liu and Fei Sun and Qi Cao and Huawei Shen},
booktitle = {AAAI 2026},
year = {2026}
}