Pre-print, Working paper: We study and compare two cases of uncertainty which have led to traditional arguments for less activist economic policies. One, as in Brainard (1967), where the uncertainty concerns the global effect of policy; and the other, as in Friedman (1960) for monetary policy, where the uncertainty is about the the lags of the effect of policy. We show that a bayesian approach of uncertainty (which is used by Brainard) leads to similar results in both cases, but that it is only in the case of uncertainty about lags that an approach in terms of robustness (through a minimax criterion) necessarily leads to less activism. Moreover, in that case, the robust policy is even less activist than what a bayesian approach would give. Therefore, a robustness criterion gives its full weight to Friedman's argument relying on uncertainty about the lags of policy.