Compounding Injustice: History and Prediction in Carceral Decision-Making

Abstract

Risk-assessment algorithms in criminal justice put people’s lives at the discretion of a simple statistical tool. This thesis explores the ways in which algorithmic decision-making in criminal policy can exhibit feedback effects, where disadvantage accumulates among those deemed ‘high risk’ by the state. Evidence from Philadelphia suggests that risk – and, by extension, criminality – is not fundamental or in any way exogenous to political decision-making. Using court docket summaries from Philadelphia, we find evidence of a criminogenic effect of incarceration, even controlling for existing determinants of ‘criminal risk’. Evaluating Philadelphia’s newly proposed sentencing tool, we suggest that algorithmic sentencing may codify and entrench existing injustice. A close look at the geographical and demographic properties of risk calls into question the use of any type of prediction in criminal policy. Finally, using probabilistic models, we explore the theoretical implications of repeated algorithmic decision-making.