Microwork as a development project: An ethnographic study of data annotators in Guizhou, China

Abstract

This paper adopts an ethnographic approach to explore microwork as a development project, focusing on the dynamic relations between the state, labor recruiting agency and workers. The Chinese state has made big investments to turn the poor and remote region of Guizhou into a big data hub, laying high hopes for high-tech to contribute to poverty alleviation. Soon the big data industry attracted the concentration of data annotation firms that vowed to train unskilled rural residents to work. We present the case of G Firm, a “complementary organizations to algorithms” (COTA) that conducts data annotation for AI platforms and meets the government’s demand for job creation. Conventional research on microwork largely focuses on how platforms such as AMT and Clickfarm exploit labor, but has paid little attention to the role of outsourced agencies in taking up the tasks of labor training and management. This paper looks at how G Firm offered a space of worker copresence to facilitate the social learning of labelling skills. However, whether annotation work is qualified or not is decided less on annotator’s individual embodied experience or peers’ social expertise than on the requirement of the inspectors. Therefore, COTA serves as an intermediary for the coding elites to exert indirect control over the cybertariat who often have to endure unpaid work due to the fast iteration process of AI. However, the fast turnover rate and fragmented division of labor made them difficult to build solidarity and assert better labor rights. Although data annotators can accomplish tasks that algorithms fail to do, given their lack of solidarity, their skills have not endowed them with high bargaining power. Our study has demonstrated the indispensable role of human labor to technological growth and would like to call for development studies to take into consideration the central role of labor as an agency for change.