The selection of optimal inputs and outputs from larger candidate sets is an important prerequisite step to enable an effective design of a robust control law. This is especially true for active damping of complex and high-dimensional multi-input multi-output systems (MIMO) with flexible modes. Existing optimal selection strategies are commonly based on a single nominal system model and cannot explicitly consider parameter variations in the system. Such nominally optimal selection, however, may perform poorly for a different parameter setting and thus deteriorate achievable robust performance. The proposed integrated input-output (I/O) selection strategy overcomes this problem by explicitly considering a set of systems covering the range of relevant parameter variation. It is structured in three consecutive selection steps, where only the subset of accepted candidates is fed into the subsequent selection; thus, it is also computationally efficient. An additional advantage is a quantitative ranking of the final candidates, which enables the control engineer to make an optimal I/O choice. The method is illustrated by a flexible beam example as well as a blended wing-body passenger aircraft model.