Agent Based Models
Agent based models (ABMs) are computational simulations of multiple agents, objects and their interactions, which aim to represent the dynamics and outcomes of system evolution. They are embedded as a key tool for complex systems explanation, exploration and partial-prediction. ABMs have strong foundations in cybertics, cellular automata and object oriented programming, and are related to agent modelling in specific domains such as ACE (Agent-based computational economics) IBM (individual based modelling in sociologically, etc).
ABMs have many strengths, including:
- the ability to create simple high-level (abstracted) representations of systems or very detailed representations;
- use of statistical distributions of populations to avoid unwieldy data collection or they can use actual data, such as time-series data of energy consumption;
- the models can be run 10s and 100s of times, each time creating unique, very slightly different initial conditions, which can lead to very different outcomes, and so enable confidence intervals to be generated about the future accuracy of 'predictions';
- use under different scenarios and provided agent rules sufficiently describe behaviours under various conditions then the survival of systems can be gauged under different futures;
- interventions, whether they are policies, regulations, rules, technologies, assets, or others, can be modelled to investigate the likely impact of change;
- identify transition paths towards desired futures, such as low carbon economies.
Like every method, ABMs are still being developed and improved. There are exciting times ahead.