This paper investigates data repair practices through a six-month-long ethnographic study in Bangladesh. Our interviews and field observations with data repairers and related stakeholders found that, alongside the scarcity of high-precision machinery and access to advanced software, data repair work is constrained by cross-language learning resources and the protective nature of documenting, curating, and sharing the experiences and knowledge among local peers. Repairers turning to external resources such as foreign forums and LLMs also revealed their frustrating experiences and the postcolonial ethical tensions they encountered. We noted that both anticipated technical labor and the emotionality of data were taken into account for pricing the data repair job, which contributed to their market sustainability strategies. Engaging with repair, infrastructure, and data poverty discourse, we argue that data repair practices represent a crucial challenge and opportunity for HCI in advancing global efforts toward data equity.
ACM CHI Conference on Human Factors in Computing Systems