Abstract
The main problem is that one cannot build a good economic model as a collection of modules prepared in advance. Each successful model is a new look on the whole economic system. It emphasizes some aspects and neglect others. Not only the set of modules and form of equations vary from model to model but the list of variables and their interpretations vary as well.
For this reason customary economic terms like ``price", ``product", ``labor", ``capital" can not be used as a base of knowledge representation in the field of economic modeling. Concepts used in economy are not collections of species but rather mappings that put into correspondence different situations, aspects and models. If we could grasp the analogies expressed by economic terms we might turn collection of different models into consistent adequate description of economy.
To solve the problem of knowledge representation we have worked out a special system of classification (canonical form) of model equations and variables. The canonical form reflects the model structure but not its economic interpretation. The main divisions of canonical form are ``agent", ``interaction", ``asset", ``plan" and ``message". ``Agent" is a set of equations which describes a decision making process and determines ``plans". ``Interaction" is a set of equations which determines ``messages" and changes ``asset" stocks according to ``plans" of different ``agents".
Rules of correctness of classification are necessary and not sufficient. One has enough freedom to relate a given item (statement or variable) to one division or another. In doing so one assigns some meaning to the item. So the classification system serves as a tool of knowledge representation.
There are some groups of morphisms of canonical forms which can help to
find similarity of parts of different models. There were also developed
methods to visualize models, to call information and to carry out
computer experiments.