Overview

The customizable knowledge capture environment is called a Discovery Machine (DM) Custom Console. A DM Custom Console is a specialized Graphical User Interface (GUI) that is tailored to a narrow set of experts. We tailor the GUI by integrating the environment into the expert's work setting. Experts do not think in terms of programming or building software systems - they think about the activities that make them experts. The DM Custom Console allows them to think about their problems as they always do.

The following are two example of DM Custom Consoles:

• Affordable Human Behavior Modeling (AHBM) - This DM Custom Console is a simplification of the standard Discovery Machine Modeler for modeling behaviors which are used to control artificial units used in military training simulations.
• BattleBuilder - This DM Custom Console customizes and extends the DM Modeler by adding a map and timeline for modeling military exercises at an abstract level.

Using a DM Custom Console

A DM Custom Console is customized through domain specific classes, enumerations, data sources, critics, strategies, procedures, functions, and user interface components. A notional DM Dashboard is shown here.

The primary input mechanism for the expert is the hierarchy building window in the center of the screen. Decompositional hierarchies are how experts encode process strategies. When a hierarchy is completed it can be run against data sets acting as a valuable assistant to the expert. However, the expert does not need to be bothered with the details of getting the hierarchy to run, only with capturing the salient aspects of the strategy. The user interface components can be connected to critic hierarchies which handle the creation of the lowest levels of the hierarchy. The critic hierarchies would be designed by a knowledge engineer (KE) for the particular domain of interest for the expert. For example, "button2" could run a modification critic which adds specific nodes to the hierarchy in the view. A menu option could add variables or values to elements within the hierarchy. "Button3" might walk the expert through a wizard using the blank text box on the bottom left to present messages to the expert. Additionally, the expert might be offered context sensitive and domain specific help with building the hierarchy as nodes are added.

The Benefits of Using a DM Custom Console

There is a vast set of literature indicating hierarchies are a cognitively valid way to capture knowledge. Humans have an innate ability to represent knowledge in hierarchies and understand the knowledge quickly. The easy to use style of hierarchy used in the Discovery MachineT allows users to work at any level of detail and quickly make modifications.

Tailoring the knowledge capture environment to the work environment of the expert through widgets, data bases, etc. allows the experts to capture their expertise as it is cued rather than documenting processes in a fabricated environment. For example, knowledge based systems have classically been built by interviewing experts and then encoding the knowledge independent of the expert. The expert is then expected to verify and validate the knowledge bases. More realistic and usable knowledge can be attained by letting the expert document his or her own knowledge in a natural way and by communicating the resulting executable models without any underlying syntax to get in the way.

Tailoring the knowledge capture interface to the domain of the expert has been a goal of knowledge engineering systems for some time (EXPECT, Prodigy). Past attempts have failed to recognize the importance of a unifying formalization across an enterprise or organization. It is important that the knowledge engineer (KE), expert, and the software engineer (SE) are all able to communicate in the same formal representations even if the formal representations are tailored depending on expertise.

Deployment of Process Models in Applications

Although captured expert process models could serve to assist the expert, the primary reason for capturing expert process models is to provide a software application for a different group of users - perhaps colleagues, novices, or a different organization. The opportunity arises to create applications based entirely on process models.

As elusive as knowledge capture has been for knowledge engineering, it is the leveraging of that knowledge for end-user applications that has provided the largest barrier. Applications that are built from expert process models provide the mechanism to leverage the best practices of an organization's key experts. Note that in producing applications from process models, knowledge engineers will have access to a wide range of critic models. In this way simple, easy to use applications can be provided quickly without the normal pitfalls of poorly designed systems or rapid prototyping. The generated applications will come with the benefit of usability expertise and with the stability of formal underlying processes. Thus these applications will not only be rapidly constructed but will also blend into the work processes of the end users, verified by the expertise of usability, software, and knowledge engineers.

End-user applications can be added to a library of applications to form components for reuse in future DM Custom Consoles and applications. In addition, particularly useful components can be made available to the development environment as building blocks.