Complexity, Evolution and Learning

A paradigm shift in economics is taking place. In traditional, neoclassical economics a representative agent who behaves perfectly rational has been the main working hypothesis and mathematical analysis of simple tractable models its main focus. A problem with this approach is that it requires unrealistically strong assumptions about individual behaviour, such as perfect knowledge and information about the economy and extremely high computational abilities to do what is optimal. An advantage of the neoclassical research programme, partly explaining its success, is that rationality imposed through optimizing behaviour and model consistent expectations enforces strong discipline on the modelling framework leaving no room for market psychology and unpredictable, irrational behaviour. An alternative complexity view is now emerging, based on interaction of many heterogeneous agents, whose behaviour is only boundedly rational. In this new behavioural agent-based approach, computer simulation models are the main modelling framework. An advantage is that it becomes possible to describe in detail individual behaviour at the micro level based on realistic assumptions. The Santa Fe conference proceedings Anderson et al. (1988) and Arthur et al. (1997a) contain many contributions within the complexity view. The recent Handbook of computational economics (Tesfatsion and Judd 2006) contains many chapters describing the state of the art of agent-based economics. There is however still an important problem with the bounded rationality research programme: it leaves too many degrees of freedom. There is only one way (or perhaps a few ways) one can be right, but there are many ways one can be wrong. To turn the alternative view into a successful research programme, one has to “tame the wilderness of bounded rationality”.