The Punitive Optimization Problem (POP) emerges in the discretionary decisions made by detention officers to override risk assessment recommendations about the release of low-risk immigrants and their continued detention. Specifically, this dynamic involves executive perspectives on detention policies and the discretion exercised by lower levels in implementing decisions that may override them. Punitive decisions stem from front-line officers who countermand higher-level recommendations to release low-risk immigrants. The mechanism comprises three key decisions: first, the upper echelon establishes the algorithmic rule; second, front-line officers opt to override it; and third, the upper echelon adjusts the algorithm in response to the front-line officer’s override. Since front-line officers drive the POP, it is helpful to place the POP within a principal-agent framework to analyze how the agent may undermine the principal’s directives.