Utrecht Fairness Recruitment dataset

This dataset contains the recruitment decisions of four companies over 500 candidates. For each candidate we have a few general descriptions (gender, age, sport) and a few indicators. The actual decision is also included. The data set can be used to get basic data science experience, but also to gain a deeper understanding of fairness.

Basic data science
The dataset is synthetic and it is suggested that students pick one company and build a prediction model for that company. This can be done via decision trees, logistic regression, neural networks or other methods and if you allow all indicators to be used, should give good results.

Understanding fairness
The recruitment decisions of the companies show different amounts of bias, that students can analyse. There are multiple variables for which bias can be analysed: gender, nationality, age and sport. if your model uses any of these as input values, your model might also be biased and you can analyse and perhaps correct this.

本数据集收录了四家公司在500名候选人中的招聘决策。对于每位候选人，我们提供了若干项基本信息（如性别、年龄、运动爱好）以及若干指标。实际的招聘决策结果亦包含在内。该数据集不仅有助于初涉数据科学领域的学习者积累实践经验，更有助于深入理解公平性议题。

基础数据科学
该数据集为合成数据，建议学生选取其中一家公司，并针对该公司构建预测模型。可通过决策树、逻辑回归、神经网络或其他方法实现，若允许使用所有指标，则预期能够取得良好效果。

公平性理解
各公司的招聘决策体现了不同程度的偏见，学生可以进行深入分析。存在多个变量，其偏见可被分析，包括性别、国籍、年龄和运动爱好。若模型使用其中任何一项作为输入值，模型本身可能存在偏见，学生可以进行相应的分析和可能的校正。