This paper proposes an original risk-based robust mixed-integer linear programming to global closed-loop supply chain network design based on a novel, uncertain, bi-level programming model. The results show that using the proposed approach significantly improves the solution robustness so that the standard deviation of profit, income, and cost is reduced by about 28%, 34%, and 36% on average. However, a more conservative DM, compared to an optimistic DM, faces more income cuts to achieve greater robustness. As far as the managerial implications are concerned, this research can assist the DM in identifying essential trade-offs among risk, cost, profit, income, robustness, and leanness regarding its attitude toward risk. Because a similar change in the level of risk-taking of the DM does not necessarily have the same economic consequences for pessimistic and optimistic DMs. In addition, the proposed decentralized structure dramatically improves the leanness of the network, especially for a pessimistic DM.