Intermediation: Algorithmic Prioritization in Practice in Homeless Services

Homelessness is a significant and growing crisis in the United States. In an effort to more efficiently and fairly distribute limited housing resources, jurisdictions across the US have adopted algorithmic prioritization systems to help select which unhoused people should receive resources. Given the impact of algorithmic prioritization on the lives of unhoused people, there is a need to more fully examine how these systems are implemented in practice by frontline workers such as social service workers. In this paper, we present a qualitative study that draws on interviews and artifact walkthroughs with fifteen social service workers to examine how they interacted with algorithmic prioritization systems as part of their job duties. We found that social service workers employed discretionary work practices to mediate between the rigid formats of algorithmic prioritization systems and the messy, situated realities of homelessness. We term these discretionary work practices 'intermediation' and provide four examples that illustrate how our interlocutors were able to maintain their commitment to advocacy and to express their expertise despite the automation of major aspects of their professional decision-making. These work practices, which we argue were motivated by care for clients, as well as a desire to preserve professional autonomy, lead us to conclude that discretion cannot easily be removed from bureaucratic systems and that removing discretion is not necessarily a desirable outcome.