Adaptive Learning in Evolving Task Allocation Networks1

We study task allocation in multi-agent systems. Task allocation has become a major research topic over the past years [3]. Here, we are particularly interested in allocation of tasks among firms on industrial, interfirm markets. These are traditionally studied using transaction cost economics (TCE). However, as has been widely acknowledged, TCE does not include dynamics of learning, trust and adaptation in its analytical framework, see, e.g. [2]. In our paper, therefore, we extend TCE with notions of trust and loyalty, dynamics and learning, in what we call Agent-based Computational Transaction Cost Economics, the application of the Complex Adaptive Systems (CAS) [1] paradigm to transaction cost economics. In order to be able to do so, we formulate our new theory at the level of individual agents, where these concepts live. In particular, we design an agent-based model, and our theory’s refutable hypotheses are derived not by deduction as usual, but by implementing and running our model on a computer.