Sweta6/FairHire

--- language: - en pretty_name: FairHireBench tags: - bias - fairness - hiring - llm - intersectionality - benchmark - algorithmic-fairness - ai-hiring - demographic-bias - gender-bias - racial-bias - responsible-ai - tabular license: cc-by-4.0 task_categories: - text-classification size_categories: - 10K<n<100K --- # FairHireBench: A Cross-Generational Intersectional Bias Benchmark for LLMs in Automated Hiring ## Dataset Description FairHireBench is a comprehensive benchmark comprising **10,005 candidate profile records** across **2,001 unique candidates** spanning **15 intersectional demographic groups** (5 racial/ethnic x 3 gender categories) for evaluating bias in AI-driven hiring systems. Each profile represents a mid-level software engineer candidate with the following attributes: | Column | Description | |---|---| | Groups | Unique candidate ID (Group 1-2001) | | Name | Candidate name | | Age | Candidate age | | Gender | Man, Woman, Non-binary | | Race/Ethnicity | African, Asian, European, Hispanic, American | | Years of Experience | Work experience in years | | Colleges | College tier | | Certification | Number of certifications | | Achievement/Awards | Number of achievements/awards | ## Intended Use Designed to audit and evaluate **intersectional bias in LLM-based automated hiring systems** using the Intersectional Fairness Evaluation Protocol (IFEP). ## Associated Paper **FairHireBench: A Cross-Generational Intersectional Bias Benchmark for Large Language Models in Automated Hiring** Sweta Jaishankar Ratnani, Lingyao Li, Yitian Lou, Mingyang Li, Kaixun Hua ## License CC BY 4.0 - free to use with attribution.语言： - 英语 官方名称：FairHireBench 标签： - 偏差 - 公平性 - 招聘 - 大语言模型（Large Language Model, LLM） - 交叉性（intersectionality） - 基准测试（benchmark） - 算法公平性 - AI招聘 - 人口统计偏差 - 性别偏差 - 种族偏差 - 负责任AI - 表格型数据 许可证：CC BY 4.0 任务类别： - 文本分类 样本规模： - 10K<n<100K # FairHireBench：面向自动化招聘场景中大语言模型的跨代际交叉性偏差基准测试 ## 数据集说明 FairHireBench是一款综合性基准测试数据集，包含**10005条候选人档案记录**，覆盖**2001名独立候选人**，涉及**15个交叉性人口统计群体**（5个种族/族裔类别 × 3个性别类别），旨在评估AI驱动招聘系统中的偏差。 每份档案均对应一名中级软件工程师候选人，包含以下属性： | 列名 | 说明 | |---|---| | Groups | 唯一候选人ID（编号1-2001） | | Name | 候选人姓名 | | Age | 候选人年龄 | | Gender | 男性、女性、非二元性别 | | Race/Ethnicity | 非洲裔、亚裔、欧裔、西班牙裔、美洲裔 | | Years of Experience | 工作年限 | | Colleges | 院校层级 | | Certification | 持证数量 | | Achievement/Awards | 成就/奖项数量 | ## 预期用途 本数据集旨在通过**交叉性公平性评估协议（Intersectional Fairness Evaluation Protocol, IFEP）**，审计并评估**基于大语言模型的自动化招聘系统中的交叉性偏差**。 ## 关联论文 **FairHireBench：面向自动化招聘场景中大语言模型的跨代际交叉性偏差基准测试** 斯韦塔·杰尚卡尔·拉特纳尼、李灵遥、楼逸天、李明阳、华开迅 ## 许可证 CC BY 4.0 - 可免费使用，需注明原作者。