ABSTRACT
Despite accounts of how artificial intelligence (AI) is replacing human labour, constant efforts are needed to keep AI-based automation running. In this chapter, we are particularly interested in data work that supports processes of automation. We explore an unconventional arrangement in which Finnish prisoners annotate text to produce training data for a local AI firm. The use of prison labour to train AI invites straightforward conclusions of exploitation of the marginalised. On a closer look, however, attention is drawn to local and situational variations of data labour: how high-tech development can be married with humane penal policies and low-cost labour with rehabilitative aspirations. We argue for an approach that can hold together seemingly contradictory value aims and open novel ways of exploring processes of automated decision-making. By acknowledging what is of value to the different parties involved, we can begin to see alternative paths forward in the study of automation.
