Model based analysis and design of human-machine dialogues through displays

Models or simulations of mental processes during the operation of technical systems can be applied to dialogue design, either during the system design or online as part of a support system. It is a current approach to build such human performance models in cognitive architectures such as EPIC or ACT-R/PM. From the engineering point of view, these architectures contain assured knowledge from cognitive science. The use of these modelling environments lead to better and less expensive human performance models because the person developing the model is constrained and guided by the assumptions and specifications of the architecture. The subject matter of such architectures leads to comparatively fine-grained models. As an example ACT-R/PM offers detailed mechanisms in the areas of memory and perception/action. Thus the complexity of human performance modelling is still quite high. Agimap is a new enhancement of ACT-R/PM to simplify modelling of perception in this cognitive architecture through tool support and sub-model reuse. Though agimap was developed in the area of chemical plant operation it is applicable in most other human-machine system domains with minor modifications. The paper introduces cognitive architectures as a tool for human performance modelling and presents agimap as a tool and as a concept. agimap provides a graphical editor to generate an XML model of the human-machine-system user interface. A XSLT transformation is used to produce ACT-R/PM code which can be deployed in an agimap enhanced ACT-R/PM interpreter. This tool chain for the development of cognitive simulations allows for an efficient usability comparison in the early phases of system development.