We introduce heterogeneity into a monetary policy committee by allowing the degree of model uncertainty to differ across members. It is shown that in this framework the stage at which members reach consensus matters. An aggregation protocol under which members only average policy deemed optimal from each member's point of view leads to more volatility compared to an alternative protocol in which members agree on a common worst-case scenario from which optimal policy is then derived. The reason is that inflation, output and the interest rate are convex functions of each member's idiosyncratic degree of model uncertainty. If the degree of model uncertainty becomes more heterogenous, inflation volatility falls due to more vigorous stabilization policy. The degree of heterogeneity across members is therefore an important determinant of macroeconomic volatility. Interestingly, the implications for the committee design under a min-max approach to model uncertainty are identical to those derived from a Bayesian approach.