Logic, Reasoning and A Programming Language for Simulating Economic and Business Processes with Artificially Intelligent Agents
Each agent has its own set of databases and rulebases to support the modelling of its decision making processes. Roughly speaking the databases record the facts pertaining to the agent or situation (including explicit representation of its knowledge and beliefs) and the rulebases hold the rules with which new inferences (on the database) can be made. Thus each agent has a rulebase for reasoning about its own behaviour and for learning in the sense described by Edmonds and Moss [3]. Each agent has two such rulebases: one of these determines specific actions to be taken in given conditions and an other, a meta-level rulebase, which can write the rules which are contained in the action-determining rulebase. With this architecture, agents can reason about their own rules of action.
The language can be used for modelling both synchronous and non-synchronous processes and has an interface particularly suited to the modelling of behaviour in organizations, markets and whole economic systems. A large collection of similar agents can be implemented as easily as one, with each agent inheriting common rules to determine its behaviour according to its type. Agents also have databases specific to each date, enabling the `knowledge' they represent to develop over time*1.
The language has a sophisticated interface (derived from that of its base language - Smalltalk) that allows the user to browse the contents of any rulebase or database at any time, or collate the results in many such databases. One view of this is shown in figure 2.
More details of the workings of SDML can be found in [7].
Generated with CERN WebMaker