Planning to Intervene Under Models of Time Inconsistency

Abstract

Time-inconsistent preferences harm a decision maker's ability to make optimal decisions. We consider the problem of procrastination in the setting of the seminal Gilbert-Mosteller optimal stopping problem and present a lightweight model with a simple representation in which it is easy to make sense of interventions. The scenario we work with is one in which an agent is tasked with a project with multiple stages (such as an undergraduate thesis), and we present two different models: an "all-or-die" model where the agent must complete the project, and a "partial credit" model, which allows for partial completion of the project. We explore what happens when a social planner intervenes in these models by enforcing a hard intermediate deadline. In this thesis, we derive the equations for the dynamic programming problem used to solve for rational and biased agents' expected welfare (and hence also the thresholds they should use for their optimal decision-making strategy) in various scenarios, and then analyze our models by computing utilities using uniform cost distributions. Finally, we show the existence of scenarios in which intermediate deadlines are effective for improving agents' decision-making and welfare and discuss their implications for practical applications.