Agent-based model

Agent-based modeling is a specific, individual- based method of computer-based modeling and simulation, closely linked to complex systems, multi-agent systems, evolutionary programming and cellular automata.

History

Agent-based modeling has its roots both in the modeling of cellular automata, as well as in the various areas of artificial intelligence. In comparative observation, the agent-based modeling can also be interpreted as an extension of cellular automata. It is a special case of a micro-simulation. Agent-based models are based on the theory of multi-agent systems.

Theory

In contrast to other types of modeling in agent-based modeling (for example, System Dynamics ) have many small units (agents) decision-making or action. The system behavior is due to the behavior of individual agents and is not imposed at the system level. If this leads to effects at the system level, which are not derivable directly from the decision algorithms of individuals, it is called emergence. In addition, a separate from the individual decisions system behavior can be implemented.

Two key aspects of the agent-based modeling are the ways to explicitly represent heterogeneous behavior and dependencies on other individuals.

This type of modeling is especially used when the focus of a question is not the stability of an equilibrium and the assumption that a process returns to an equilibrium, but the question of how a system can adapt to changing conditions (robustness ). Here, the knowledge is taken into account that complex problems require, the micro - level, ie the decisions of individuals, their heterogeneity, and their interactions, to investigate directly.

Application Examples

Very different applications fall within the scope of the agent-based modeling. They differ for example in the degree of intelligence of the modeled agents and in the modeling of physical or social space. All these approaches have in common that the decision behavior is implemented at the level of individuals.

Some examples illustrate this bandwidth.

Simple simulation of the formation of traffic jams

In one dimension, vehicles move ( the agents ). The drivers and vehicles have a certain acceleration and braking behavior and keep a minimum distance a to the vehicle in front of them car. The complexity of the simulated environment is so small and the necessary artificial intelligence agent limited so that is possible to speak in this case of a micro- simulation. Still can meet interesting statements with these models.

A traffic model with discretely modeled space ( the vehicles move on grid cells ) is the Nagel-Schreckenberg model, an example of a car-following model with continuous space is the Wiedemann model.

Formation of ant trails

Similarly simple enough intelligence simulated ants secrete in search of food fragrances and follow the scents of other ants. The fragrances are lost with time. The two dimensional environment can be much more expensive here already, for example, contain food sources and obstacles. Even if the behavior of individuals is simple, can form a complex swarm intelligence here. See also implemented in the NetLogo simulation of the emergence of an ant trail. From this behavior, so-called ant algorithms are derived for solving combinatorial optimization problems.

Segregation

Something more elaborate decision-making behavior show the agents in Schelling's segregation model. There, agents take account of different preferences, a choice in which district they move. To the spatial environment is the social environment here. The behavior of the agent depends on the behavior and preferences of other agents from ( social embeddedness ).

Social Networks

Space can occur entirely in the background, if the decision of the agents no longer the place where they reside, but rather on the other agents with whom they have contact, such as in consumer behavior or the spread of cultural norms. These social networks are simulated. Exchange takes place only with the agents to which a network relationship. This is where the decision behavior of individual agents may well be more complicated and complex and, for example, as in the mentioned Consumats, repetition, imitation, social comparison and reflection included.

Artificial economic systems

The scientific discipline Agent -based Computational Economics deals with the simulation of economic decision making at the level of individuals. Investigated issues range from auction behavior on individual labor input ( moral hazard) to behavior in social dilemmas.

Social simulation

The field of social simulation focuses on the modeling of concrete, observable situations that are analyzed in case studies. The resulting agent-based models represent the behavior of the people in the study areas, for example, farmers in a river catchment area, from. At the same time they can be coupled with more or less complex models of the physical environment and contain appropriate feedback. A recent example of this is: Berger, Birner, Diaz, McCarthy, Wittmer (2007): Capturing the complexity of water uses and water users within a multi -agent framework, Water Resour Manage 21:129-148

Why Agent-based modeling?

Agent-based modeling is characterized primarily by the ability to model the connections between the micro and the macro level explicitly or investigate. This aspect is required in different questions:

• When will be studied, from which type of individual decisions, the aggregate behavior, ie the behavior at the system level, is formed.
• When will examine the ways in which changes are reflected on the system level (possibly heterogeneous ) behavior of individuals.
• If the combination of these two mechanisms is investigation.

Agent-based modeling and economics

The application in Artificial Economics is especially noteworthy, because the assumption of rational acting individuals (homo economicus ) was always a description at the aggregate level. The aggregate behavior of economy individuals can be described as if the individuals act rationally. For markets with a lot of information, many learning opportunities, plenty of time and motivation, this may be true. But there are plenty of examples of situations where supply assumptions of rational behavior not good forecasts of actual human behavior. The interesting scientific questions, especially in relation to public goods and social dilemmas are among these situations. Since there is no other theory about human behavior, which is suitable in the same way for aggregation, such as the rationality, it is necessary to such questions, to investigate the heterogeneous, actually observable behavior of people. Agent-based modeling is a method to simulate this behavior and build hypotheses about the relationships between the micro - behavior of individuals and the macro - behavior of the system and to investigate.