Algorithmic Interventions: The Power and Politics of Algorithmic Decision Systems

Abstract

Algorithmic Decisions Systems (ADS) are now commonly integrated within governing institutions ranging from the criminal justice system to social welfare programs. In an apolitical world, we might expect these considerations to be purely decided on bureaucratic optimization, but within highly politicized and contentious policy areas, we can expect each of these decisions to be opportunities for strategic interactions between interested parties. My dissertation seeks to address three core questions regarding governmental adoption of Algorithmic Decision Systems. 1) What attributes of these systems may lead legislators to support their use? 2) Does the inclusion of ADSs increase legislative support for bills that include these systems? 3) Do changes in the narrative around ADSs, particularly perceived public backlashes, impact legislative support? Throughout Chapters 1 and 2, I trace the history of ADSs as a natural evolution of bureaucratic systems and unpack the characteristics of ADSs that may be attractive to policymakers. In Chapter 3, I use a game-theoretic model to explore the way that ADSs are used to expand legislative control over bureaucratic decision-making. In Chapter 4, I use an empirical model to analyze legislative support for criminal justice legislation in U.S. state legislatures over the period 2012-2018, I provide evidence that suggests that inclusion of ADSs did increase some forms of legislative support for bills that included them, but that these effects were eroded and then overcome in later years by the recent critical turn and public backlash against ADSs. Lastly, I conclude my dissertation by discussing possible research avenues for future scholarly work on governmental adoption of Algorithmic Decisions Systems.