This paper introduces a new model for decision making under ambiguity called second-order uncertainty. The decision maker considers all probability distributions associated with the given options, and chooses the allocation that minimizes the variability of outcome across all probabilities. Optimal allocation under second-order uncertainty is the equal weighted allocation, also called naive diversification.