Adaptive User Interfaces are seen as a critical success factor in the development of training and learning systems. Adaptive interfaces have been based on an approach consisting of user and device profiles. Recently, personality and mental states have been added and are used in research projects to expand the reliability and context awareness of such systems. This approach enhances adaptive usage of training and therapeutic systems. The developed system effectively combines biofeedback sensors and a set of software algorithms to estimate the current motivation/frustration level of the user. Based on this concept, it will be possible to develop narrative training and therapeutic systems, which could adapt to the motivation level of the user and focus her attention on the fulfilment of the current task.