ABSTRACT
This chapter traces what becomes visible and invisible through practices in the AI supply chain such as data labelling. The chapter asks how a machine knows who is a woman, a man or non-binary. The chapter shows how computers see the world through labels that are often assigned by humans. This human labour in the AI supply chain is regularly made invisible. The research shows how such work is often conducted in the Global South. While this work is regularly seen as problematic in regard to working conditions, a central interest of the chapter is to explore the construction processes that happen in regard to data practices. The chapter shows that labels applied to data are conceived by designers of AI and applied by data labellers, which is a subjective process. It functions to construct subjective ways of seeing the world as universal and objective. The chapter argues that processes of data labelling have to be made visible and tangible to allow for inclusion.
