Models with heterogeneous interacting agents

According to David Colander, economic research in the XXI century has entered post-neoclassical era. The domain of neoclassical orthodoxy is over, and advanced Economics is looking for alternatives.
In my opinion, re-foundation of Economics requires a fundamental step: homo economicus social agent who learns, abandoning the mechanistic method of physical reductionist, and acknowledging that economic agents are atoms that think and learn.
Agent-based models (ABM) have potential to meet the challenge, but need additional development to offer new and comprehensive paradigm. ABM are abstract representations of reality characterized by:
– multiplicity of agents interacting with each other and with the environment,
– autonomy of agents, i.e., no central regulator on agents’ behavior, and
– interactions can be calculated numerically and analytically studied.
In other words, ABM are instruments used to study the economy as complex adaptive system, using as basic units heterogeneous agents, ranging from individuals to organized social groups that perform economic tasks and may also be aggregated to other agents to form industries or countries.
The balance of an economic system no longer requires each individual element in equilibrium, but rather that statistical distributions that describe the aggregation phenomena are stable. ABMs were used to analyze macro-economy, crises and cycles, distribution of wealth, income and company size, financial markets and the way market structures respond to technological change. A further field of development of many ABM has been the study of dynamics of learning and behavioral experimental approach. Finally, another area of interest for application of ABM is the analysis of networks.