Towards Ethical Outsourcing of Data Work in Academia

There is a surging demand for data work, i.e., the labor involved in collecting, curating, annotating, and verifying data, which is required to make data available, readable, and analyzable to machines. Companies and research institutions that rely on this work to create so-called AI products or conduct research—the data work requesters—typically outsource this labor. Previous research has investigated various types of data work that are important for research yet under-recognized by scientific communities. In addition, several initiatives have addressed data work outsourcing in industry settings. However, academic outsourcing practices remain largely understudied, with key decisions about data working conditions often left to individual researchers and teams. Based on two participatory workshops and two focus group discussions between data labor researchers, academic requesters, and data workers, this paper examines the challenges and needs of data workers and academic requesters and explores ways to address them in academia. We expand existing research by providing practical recommendations and developing a tool that assists academics with writing task instructions. Through this intervention, we aim to support academics and IRBs to oversee and ensure ethical practices in data work outsourcing.