当前位置: X-MOL 学术arXiv.cs.IR › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Unintended Bias in Language Model-drivenConversational Recommendation
arXiv - CS - Information Retrieval Pub Date : 2022-01-17 , DOI: arxiv-2201.06224
Tianshu Shen, Jiaru Li, Mohamed Reda Bouadjenek, Zheda Mai, Scott Sanner

Conversational Recommendation Systems (CRSs) have recently started to leverage pretrained language models (LM) such as BERTfor their ability to semantically interpret a wide range of preference statement variations. However, pretrained LMs are well-known to be prone to intrinsic biases in their training data, which may be exacerbated by biases embedded in domain-specific language data(e.g., user reviews) used to fine-tune LMs for CRSs. We study are recently introduced LM-driven recommendation backbone (termedLMRec) of a CRS to investigate how unintended bias i.e., language variations such as name references or indirect indicators of sexual orientation or location that should not affect recommendations manifests in significantly shifted price and category distributions of restaurant recommendations. The alarming results we observe strongly indicate that LMRec has learned to reinforce harmful stereotypes through its recommendations. For example, offhand mention of names associated with the black community significantly lowers the price distribution of recommended restaurants, while offhand mentions of common male-associated names lead to an increase in recommended alcohol-serving establishments. These and many related results presented in this work raise a red flag that advances in the language handling capability of LM-drivenCRSs do not come without significant challenges related to mitigating unintended bias in future deployed CRS assistants with a potential reach of hundreds of millions of end users.

中文翻译:

语言模型驱动的会话推荐中的意外偏差

会话推荐系统 (CRS) 最近开始利用预训练语言模型 (LM),例如 BERT,因为它们能够在语义上解释各种偏好声明变体。然而,众所周知,预训练的 LM 在其训练数据中容易出现内在偏差,这可能会因嵌入在用于为 CRS 微调 LM 的特定领域语言数据(例如,用户评论)中的偏差而加剧。我们研究了最近引入的 CRS 的 LM 驱动的推荐主干(称为 LMRec),以调查无意的偏见,即语言变化,如姓名参考或不应该影响推荐的性取向或位置的间接指标,如何体现在价格和类别分布的显着变化中餐厅推荐。我们观察到的令人震惊的结果强烈表明,LMRec 已经学会通过其建议来强化有害的刻板印象。例如,随口提及与黑人社区相关的名称会显着降低推荐餐厅的价格分布,而随口提及与男性相关的常见名称会导致推荐的酒精服务场所增加。这项工作中提出的这些和许多相关结果提出了一个危险信号,即 LM 驱动的 CRS 的语言处理能力的进步并非没有与减轻未来部署的 CRS 助手中可能达到数亿终端的意外偏见相关的重大挑战用户。随口提及与黑人社区相关的名称会显着降低推荐餐厅的价格分布,而随口提及与男性相关的常见名称会导致推荐的酒水服务场所增加。这项工作中提出的这些和许多相关结果提出了一个危险信号,即 LM 驱动的 CRS 的语言处理能力的进步并非没有与减轻未来部署的 CRS 助手中可能达到数亿终端的意外偏见相关的重大挑战用户。随口提及与黑人社区相关的名称会显着降低推荐餐厅的价格分布,而随口提及与男性相关的常见名称会导致推荐的酒水服务场所增加。这项工作中提出的这些和许多相关结果提出了一个危险信号,即 LM 驱动的 CRS 的语言处理能力的进步并非没有与减轻未来部署的 CRS 助手中可能达到数亿终端的意外偏见相关的重大挑战用户。
更新日期:2022-01-19
down
wechat
bug