Market agents with a sense for mechanisms

When software agents act on markets they typically face a predetermined market mechanism. Their strategies are a priori tailored by the agent designer to suit the market mechanism at stake. So, the market mechanisms are only considered and examined at design time. In this paper, we present an approach to how agents themselves can grasp market mechanisms at runtime, directly derive a valid interaction protocol and consider the mechanism structure for their strategies - even for mechanisms they have not known before. For exemplifying our ideas, we present a BPEL implementation of a reverse English auction that is based on the notion of an auction reference model and make the BPEL description available to the agent. We show how the agent will derive an interaction protocol - the valid behavior - and how it even forms a reasonable strategy when presuming some generic market knowledge and basic strategy building blocks