This paper considers the performance of average inflation targeting (AIT) policy in a New Keynesian model with adaptive learning agents. Our analysis raises concerns regarding robustness of AIT when agents have imperfect knowledge. In particular, the target steady state can be locally unstable...