The optimal control approach to monetary policy has garnered increased attention in recent years. Optimal control policies, however, are designed for the specific features of a particular model and therefore may not be robust to model misspecification. One important source of potential misspecification is how agents form expectations. Specifically, whether they know the complete structure of the model as assumed in rational expectations or learn using a forecasting model that they update based on incoming data. Simulations of an estimated model of the U.S. economy show that the optimal control policy derived under the assumption of rational expectations can perform poorly when agents learn. The optimal control approach can be made more robust to learning by deemphasizing the stabilization of real economic activity and interest rates relative to inflation in the central bank loss function. That is, robustness to learning provides an incentive to employ a "conservative" central banker. In contrast to optimal control policies, two types of simple monetary policy rules from the literature that have been found to be robust to model misspecification in other contexts are shown to be robust to learning.