Abstract
Jurisdictions increasingly employ algorithms in public sector decisionmaking. Facing public outcry about the use of such technologies, jurisdictions have begun to increase democratic participation in the processes by which algorithms are procured, constructed, implemented, used, and overseen. But what problem is the current approach to democratization meant to solve? Policymakers have tended to view the problem as the absence of public deliberation: agencies and courts often use algorithms without public knowledge or input. To redress this problem, jurisdictions have turned to deliberative approaches designed to foster transparency and public debate. Jurisdictions have begun to increase democratic participation in the processes by which algorithms are procured, constructed, implemented, used, and overseen. But what problem is the current approach to democratization meant to solve? Policymakers have tended to view the problem as the absence of public deliberation: agencies and courts often use algorithms without public knowledge or input. To redress this problem, jurisdictions have turned to deliberative approaches designed to foster transparency and public debate.
This Article contends that the current approach to democratization is too narrow a solution if we seek to redress how algorithms mediate the political powerlessness experienced by oppressed groups. The problem with algorithms is not the mere absence of public input; it is one of power. Current algorithms operate to entrench state practices that suppress the democratic participation of oppressed groups, reinforcing their economic and social inequality as well as their structural marginalization in governance. While deliberative approaches have attracted broad political support, these reforms cannot reach these deeper power concerns. Moreover, the current approach to ‘democratization’ comes with a hidden cost: by building trust and legitimacy around algorithms, it can distract attention from how the state’s current use of algorithms exacerbates existing inequalities, power imbalances, and social stratification.
Addressing these harms may be possible, but doing so requires equalizing how power is distributed among different groups within our current democratic institutions and our society more generally. In this sense, the democratic participation problem facing algorithms extends beyond the challenges of regulating a new technology; instead, the problem represents a microcosm of oppressed people’s broader struggle for full participation in this country’s democracy. Building a more inclusive democracy will be a long-term, and ongoing, project. As we struggle to move toward that world, the aim should be to build processes that endow oppressed groups with the power to resist algorithmic technologies that reinforce their political, economic, and social subordination in the current moment.
