Agent Based Modelling

Agent-based geo-simulations are a set of geo-simulations based on object-oriented principles and implemented in an object-oriented programming language.

Limitations

• a model is only as useful as the purpose for which it was contracted. A model has to be built at the right level of description for every phenomenon.
• Payoff of bottom-up models versus their complexity. Payoff is determined by its structural realism; i.e. its ability to produce independent predictions that match observations.

Components of Geo-simulation:

1. Agent

A. Characteristics of Agent:

According to Macaland North, Agent is identifiable (discrete), goal-directed, autonomous and self-directed, flexible (can learn and adapt its behavior), situated in an environment and interact with it.

Meanwhile, Nigel Gilbert explained agent has perceived their environment, and performance-behavior (motion-action-communication), has a memory and can determine which behavior to perform.

B. Behaviour of Agent: 

There are two types of behavior:

• Internal behavior (where the agent does not need external solicitations).

• External behavior (when receiving external solicitations such as events).

C. Learning Pattern of Agent:

Agents can learn in a supervised and in an unsupervised way, both individually or in groups.

D. Typology:

Typology is an artificial way to define different groups based on specific criteria to organize and analyze reality (McKinney, 1950; Jollivet, 1965).

In this case, agent typology can be designed (i.e. provided with simplified characteristics) or analyzed (empirical – representing an actual population).

2. Environment

"Environments define the space in which agents operate, serving to support their interaction with the environment and other agents"

"The environment can be coinsidered as the world in which the agents occupy and is often represented in a patchwork. Each patch has parameters which affect how the agents interact with it, for example, a grassland environment where some patches are occupied by prey animals." (kilian and simone, 2020)

3. Environment - Agent interaction  

Figure bellow shows the simplest environment- agent-agent interaction in the of ABM . Agents derive information from the environment that informs the perception they have about the state of the environment. Based on the goals and attributes of the agents they make decisions on actions to perform and these actions affect the environment. The agents can interact indirectly, for example by affecting the common resource, or directly by communication. This communication might be used to exchange information about possible strategies, knowledge about the resource and agreements how to solve collective action problems (Marco, 2005).

https://discovery.ucl.ac.uk/id/eprint/3342/1/3342.pdf

3. Time

Limitation of Agent based Modelling

The enthusiasm of adopting the ABM approach for modelling geographical systems is curtailed by some limitations. Although common to all modelling techniques, one issue relates to the purpose of the model; a model is only as useful as the purpose for which it is constructed. A model has to be built at the right level of abstraction for every phenomenon.

If the level of abstraction is too simple, one may miss the key variables. Too much detail, and the model will have too many constraints and become overly complicated. The nature of the system being modelled is another consideration. For example, a system based on human beings will involve agents with potentially irrational behaviour, subjective choices, and complex psychology. These factors are difficult to quantify, calibrate, and sometimes justify, which complicates the implementation and development of a model, as well as the interpretation of its simulation outputs. However, the fundamental motivation for modelling arises from a lack of full access to data relating to a phenomenon of interest. Often, the target itself is neither well-understood nor easy to access.

Application of ABM

ABMs have been developed for a diverse range of subject areas, such as: archaeological reconstruction of ancient civilisations ; understanding theories of political identity and stability; understanding processes involving national identity and state formation; biological models of infectious diseases; growth of bacterial colonies; single- and multi-cellular level interaction and behaviour; alliance formation of nations during the Second World War; modelling economic processes as dynamic systems of interacting agents; company size and growth rate distributions; geographical retail markets, size-frequency distributions for traffic jams; price variations within stock-market trading; voting behaviours in elections; identifying and exploring behaviour in battlefields; spatial patterns of unemployment; trade networks ; business coalitions over industry standards; social networks of terrorist groups.

Validation of ABM

Validation is the process of determining the degree to which a model or simulation is a reliable representation of the target system, or “real world,” from the perspective of the intended uses of that model or simulation. The validation process is the process of comparing the model outcome with its referents and its validation data in order to evaluate the model’s accuracy. Potential referents exist in many forms, varying from subjective and qualitative descriptions to objective and quantitative descriptions:

•  Experimental data describing the functionality and performance of a system. • Empirical data describing the behavior of a system
•  Experience, knowledge, and intuition of experts • Mathematical models of the behavior of a system
•  Qualitative descriptions of the behavior of a system
• Combinations of the types described above The process of checking whether or not a model is consistent with respect to some formalism or theory is known as verification.

Verification determines whether the design and implementation of a model or simulation correctly meet the design requirements as best reflected in a validated conceptual model. Experimental Data Describing the Functionality and Performance of a System For the simpler models, it is possible to conduct Monte Carlo uncertainty analysis. When constructing such models, but probably also the more complex (many parameters, complicated decision processes with group influences, etc.), the sensitivity of the various aspects and components will be analyzed.

Empirical Data Describing the Behavior of a System The model will be validated (and calibrated) with:

• GIS maps created and derived from satellite images/remote sensing
•  Participatory perceived-value maps derived via interviews • Available census and socioeconomic data
• Data retrieved via household surveys
•  Data retrieved via literature reviews and from archives.