FairJob
收藏资源简介:
FairJob数据集由Criteo AI Lab和Université Paris Dauphine-PSL创建,专注于在线职位推荐广告的公平性研究。该数据集包含大量匿名用户上下文和发布者特征,来源于为期5个月的职位定向广告活动。数据集通过非均匀子采样和特征随机投影进行处理,以保护商业机密并防止原始特征或用户上下文的恢复。尽管缺乏明确的敏感属性,数据集通过代理属性保持了预测能力,适用于探索广告过程中的不公平性及其缓解技术。
The FairJob dataset was created by Criteo AI Lab and Université Paris Dauphine-PSL, focusing on fairness research for online job recommendation advertisements. It includes a large number of anonymous user contexts and publisher features, which are derived from a 5-month job-targeted advertising campaign. The dataset was processed via non-uniform subsampling and random feature projection to protect trade secrets and prevent the recovery of original features or user contexts. Although it lacks explicit sensitive attributes, the dataset retains predictive power through proxy attributes, making it suitable for exploring unfairness in the advertising process and corresponding mitigation techniques.




