Procrastination and Learning About Self-Control

We analyze conditions in which an agent does not learn about self-control problems over time in a simple model of task completion. While the agent is initially uncertain about own future self-control, in each period she has an opportunity to learn about own future self-control by paying a non-negative learning cost. If the agent is time-consistent, she always chooses to learn whenever the learning is beneficial. If the agent has time-inconsistent preferences, however, she may procrastinate such a learning opportunity. We show that the agent may procrastinate it even under zero learning cost, if the learning takes time and her preference exhibits an inter-temporal conflict between future selves (e.g., hyperbolic discounting). When the agent has multiple initially-uncertain attributes (e.g., own future self-control and own ability for the task), the agent's endogenous learning decisions may be misdirected – she chooses to learn what she should not learn from her initial perspective, and she chooses not to learn what she should.