WP4 Economic experiments will be used to calibrate agent decision-making. These experiments will study the interaction of behavioural rules and the emergent aggregate phenomena they co-create in complex nonlinear macro- and financial feedback systems. Taylored experiments to capture key situations that played a critical role in the crisis will also be designed.

While economic experiments are now a well accepted method in economics, we know of no prior studies where this has been used in close conjunction with the construction of an agentbased model, in which the experiments are designed around the needs of the model and the model is constructed based on the results of the experiments.

The basic ground work to do this has already been laid by the Hommes group in Amsterdam, who has pioneered the interpretation and analysis of economic experiments in terms of dynamical systems models. Individual expectations about future economic variables, such as asset prices or inflation, play an important role in the economy. Expectations about the future shape economic decisions of consumers, firms and investors and, hence, affect today's realizations of aggregate variables. In turn, these realizations affect individual expectations.

One may think of the economy as a complex expectation feedback system. The standard approach to modelling agents' expectations has been to assume rational expectations, when market participants form model-consistent predictions of future outcomes, as if they have full knowledge of the economic law of motion. However, rational expectation models are not satisfactory, neither in their predictions of the economy nor in their highly unrealistic assumption of full rationality. For example, the strong recent decline of financial markets and the current economic crisis are hard to reconcile with the rational model. Moreover, nonrational, heterogeneous expectations are frequently found in survey data.

There exist many different agent-based simulation models with boundedly rational agents, heterogeneous expectations and learning. Many of these share the realistic feature that they can mimic the stylized facts in financial time series and in macro data. However, the results from numerical simulations of these models are critically dependent upon which behavioural decision rules are assumed. There are simply too many ways of modelling non-rational behaviour. In order to discipline this so-called "wilderness of bounded rationality'' this WP will study, by running laboratory and internet experiments with paid human subjects, how people behave in complex nonlinear feedback environments.

The WP will pursue the following main goals:

• To study, by means of laboratory and large scale internet experiments with paid human subject, behavioural decision rules (e.g. individual expectations) and aggregate behaviour in complex financial and macro-economic nonlinear feedback environments;

• To develop a behavioural model of expectation formation and bounded rationality consistent with the experimental data, that can be used as a building block in large-scale agent based models of financial and macro-economic crises in other WPs.