2.2.4 AGENT MODELS

Systematic design can make use of agents devoted to specific phases of conceptualization, implementation, and evaluation within the ASE design process model. It has been shown how the communication between these agents can be manifested as critiques, advice, or the design artifact itself.

The inter-agent nature of the ASE model and the sharing of results through the design model has been explored, but here also exits models specific to the agents themselves. It has been observed that the models used by the ASE agents are likely to be distinct and independent, each specializing in a purpose.Although previously implied, Figure 11 makes explicit the types of models used by these agents. Evaluation, synthesis, and analysis can be defined as the processing of one model, that of design, in terms of another.

Figure 11 ASE Agents Require Models of Their Own

Evaluation agents use what will be defined here as behavior models. These may either represent the ideal space of design or the constraint-detection space. One type of behavior model defines some bounded, potential design space in which a design may properly exist. The constraint-detection model defines the space in which the design is not permitted to stray. With these models the evaluation agent can declare where the design respectively left or entered the spaces of the evaluation model.

Analysis agents use technique models to synthesize a viable list of design alternatives, while synthesis agents base integration of advice on a decision model.

It is possible that the same model is used for more than one class of agent. For example, the evaluation model which detects constraint violations may be sufficient in acting as technique model in suggesting a design change. However, in terms of better effectiveness and maintenance, models use representations and operators that are distinct across classes.