Data Donation: A First Step Towards Improving Representativeness in Algorithmic Hiring Datasets

The use of algorithmic hiring is on the rise, becoming more and more common due to its capacity to process a large number of applications with efficiency. Despite the advantages of this approach, there have been numerous cases of discriminatory hiring outcomes. A significant contributing factor to such outcomes is the lack of representativeness in the data used to develop these systems. This results in a significant decline in performance for underrepresented groups, disproportionately impacting marginalized communities. Addressing bias in algorithmic hiring requires access to comprehensive datasets that include curriculum vitae (CV) and demographic information reflecting diverse backgrounds. Unfortunately, there is a lack of datasets that serve such a purpose. This paper introduces a data donation campaign designed to collect real-world CVs, including demographic and sensitive information, from a representative sample of individuals in the EU workforce. The paper discusses the design decisions underpinning the campaign, along with the challenges encountered during its deployment and execution. Finally, it offers lessons learned and practical solutions to overcome these challenges, thereby contributing valuable insights for future efforts in this domain.