一句话：提出自动化框架 BiasAsker，通过构建社会偏见数据集并生成问题，识别和度量对话式 AI 系统中的绝对偏见与相关偏见。
对话式 AI 系统（如 ChatGPT、Siri）已广泛部署，但其数据驱动的黑盒特性使其可能输出带有偏见与刻板印象的内容，带来社会问题。现有偏见检测面临两大困难：一是缺乏同时包含社会群体与偏见属性的数据，难以生成能全面触发偏见的输入；二是现代对话系统能产生闲聊、解释等多样回复，仅基于情感与毒性的方法难以适用。本文提出 BiasAsker 自动化框架：首先构建包含 841 个社会群体和 5,021 个偏见属性的综合社会偏见数据集，然后自动生成问题，并采用基于存在性度量的新方法识别两类偏见——绝对偏见与相关偏见。在八个商用系统和两个著名研究模型（如 ChatGPT、GPT-3）上的实验表明，BiasAsker 生成的问题中有 32.83% 能触发这些系统的偏见行为。代码、数据与实验结果均已公开。
Powered by advanced Artificial Intelligence (AI) techniques, conversational AI systems, such as ChatGPT, and digital assistants like Siri, have been widely deployed in daily life. However, such systems may still produce content containing biases and stereotypes, causing potential social problems. Due to modern AI techniques’ data-driven, black-box nature, comprehensively identifying and measuring biases in conversational systems remains challenging. Particularly, it is hard to generate inputs that can comprehensively trigger potential bias due to the lack of data containing both social groups and biased properties. In addition, modern conversational systems can produce diverse responses (e.g., chatting and explanation), which makes existing bias detection methods based solely on sentiment and toxicity hardly being adopted. In this paper, we propose BiasAsker, an automated framework to identify and measure social bias in conversational AI systems. To obtain social groups and biased properties, we construct a comprehensive social bias dataset containing a total of 841 groups and 5,021 biased properties. Given the dataset, BiasAsker automatically generates questions and adopts a novel method based on existence measurement to identify two types of biases (i.e., absolute bias and related bias) in conversational systems. Extensive experiments on eight commercial systems and two famous research models, such as ChatGPT and GPT-3, show that 32.83% of the questions generated by BiasAsker can trigger biased behaviors in these widely deployed conversational systems. All the code, data, and experimental results have been released to facilitate future research.
DOI 原文 · arXiv · PDF（开放获取） · DBLP
@inproceedings{WanWHGBL23,
title = {BiasAsker: Measuring the Bias in Conversational AI System},
author = {Yuxuan Wan and Wenxuan Wang and Pinjia He and Jiazhen Gu and Haonan Bai and Michael R. Lyu},
booktitle = {FSE},
year = {2023},
doi = {10.1145/3611643.3616310}
}