Multi-Agent Architecture for Knowledge-Driven Decision Support - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 42

Figure 9. Decision tree representation of the rule-based knowledge module for the Play Tennis Problem
in XML format
day given the conditions entered by the user. The user is given information about each of the atmo-
spheric conditions and their categories. These atmospheric conditions form the domain attributes for
the play tennis problem and define the context specific information that is pertinent to this decision
problem. The agent provides information on each domain attribute thereby informing the user through
the process of selection of the attributes that are pertinent on any given day. The representation of the
domain attributes generated by the agents shows the DTD and the XML files (see Figures 5 and 6) for
the representation of information about the context of the problem domain. The XML representation of
the domain attributes is dynamically parsed by the user agent to generate a context specific user interface
(as shown in Figure 7). This allows the user to make a decision about each pertinent domain attribute.
After the user makes a selection from all the domain attributes, the user agent has enough information
to make a decision about the problem domain. This is accomplished by parsing the set of domain rules
that specify a final decision based on observations of domain attributes. As mentioned earlier, the user
agent has access to an XML representation of domain rules about a given problem context, through the
XML DOM object contained in the domain knowledge object for a decision problem. The structure
for this set of rules is shown as a DTD in figure 8, while figure 9 shows the XML representation of the
rules used by IKMDSA for the play tennis problem. The user agent parses these rules and identifies
the rules that are fired for the given set of observations. These rules are then presented to the user in
user-friendly format as explanation fro the decision made by the user agent.
In the prototype implementation of the proposed IKMDSA architecture, we use the Java program-
ming language to implement the agents as extensions of objects. The choice of Java was based upon