A knowledge hierarchy model for adaptive multi-agent systems

In a changing world the need for adaptivity in software systems is apparent more than ever, since the cost of software maintenance is a huge burden. Adaptivity is needed since business processes, business rules and business terms constantly evolve. This paper argues for a radical solution making use of the inherent adaptivity of software agents. The Adaptive Agent Model (AAM) is described in terms of a hierarchical structure for business knowledge consisting of a business concepts layer, a business rules layer and a business process layer. Collectively, these form a knowledge base that is then made available to agents running in systems. Such knowledge is sourced from the business requirements and becomes the basis for agent behaviour. Because the knowledge is externalised from the system, system behaviour is easily maintained. Making use of a case study, the paper describes how the knowledge hierarchy is made up and assesses its contribution to the goal of software adaptivity.