Plea Bargaining, Principled Sentencing, and Artificial Intelligence

Abstract The development of AI sentencing algorithms raises the prospect of more consistent and transparent sentencing outcomes. However, the hoped-for improvements in sentencing may not occur in legal jurisdictions that permit prosecutors an outsized role in determining sentencing outcomes through plea bargaining. In what I term “robust” plea bargaining regimes, prosecutors often view charging decisions and sentencing recommendations strategically. They negotiate them in ways aimed primarily at producing guilty pleas rather than producing consistent or truthful outcomes. Prosecutors in such regimes might resist efforts to make the plea deals they forge more consistent and transparent. Given the centrality of charging decisions in determining sentencing outcomes in robust plea bargaining regimes, I examine the feasibility of devising AI charging algorithms.