To physically interact with a rich variety of environments and situation-oriented requirements, humans continuously adapt both the stiffness and the force of their limbs through antagonistic muscle coactivation. Reflecting this behaviour in prostheses may promote control naturalness and intuitiveness and, consequently, their acceptance in everyday life. We propose a method capable of a simultaneous and proportional decoding of position and stiffness intentions from two surface electro-myographic sensors placed over a pair of antagonistic muscles. First, the algorithm is validated and compared to existing control modalities. Then, the algorithm is implemented in a soft under-actuated prosthetic hand (SoftHand Pro). We investigated the feasibility of our approach in a preliminary study involving one prosthetic user. Our future goal is to evaluate the usability of the proposed approach executing a variety of tasks including physical social interaction with other subjects (see Figure 1). Our hypothesis is that variable stiffness could be a compromise between firm control and safe interaction.