Model-based universal interaction devices are already capable to react on contextual changes by automatically adapting the user interface, but without considering the usefulness of the resulting user interface. Often tasks cannot be executed any more or execution orders will result in dead locks caused by unavailable functionality. We present our approach of investigating this property of adapted models based on the example of the SmartMote in our living lab the SmartFactory KL. Given the task description of the user interaction we determine a dialog model in terms of a state machine - which is necessary in our process of user interface generation - to determine possible execution orders leading to the accept state of this state machine. Using these execution orders the initial task model can be adapted, all misleading tasks can be removed and the resulting user interface will only offer valid user interactions.