Design optimization for aircraft in the presence of probabilistic uncertainties is a challenging task drawing increasing attentions in engineering fields. Based on the Bayesian inference technique, this paper aims to develop an adaptive experiment design method for further improving the accuracy and efficiency of the robust optimization design when it is applied to the time-consuming computer simulators. A probabilistic interpolation method is first used to learn the objection and constraint functions under uncertainty. The effectiveness of the proposed method is demonstrated on an aerodynamic shape optimization under uncertainty. Results of the engineering example show that the proposed method yields more robust design results than the determined optimization method.