If you want to build a bridge then engineering and materials science will give you the answers you need, all of them definitive and founded in hard evidence. If you want to prevent your citizens from becoming obese, the answers will be spread across a dozen or more disciplines, and the evidence will be of variable quality.
For organisations who are serious about understanding human behaviour, agent-based modelling is the best way to extend and improve that understanding and provide a robust basis for developing and testing strategy.
By examining and integrating all relevant, interdependent influences on a population into a coherent model, a much deeper level of causal insight is provided than is achievable by other methods
Being able to create worlds where anything can happen provides a powerful explorative tool for predicting both the range and likelihood of possible outcomes
Exploring what might happen allows us to evaluate how different strategies might perform in a myriad of scenarios allowing you to identity the optimum approach, the likely outcome and the inherent risks.
A population of agents: an agent is a representation of a real decision making unit within the real population. In most cases it is a person, but it could be a household or a company or other collective. Each agent will have a set of attributes such as age, gender and income, distributed to make the simulated population representative of the real population.
A set of rules of behaviour: each agent is provided with a set of rules that governs the choices it makes in the domain of interest, which could be where they shop, what brand of soft drink they buy, or whether they quit smoking or not. In our software we codify the rules that real people use to determine their behaviour. They are never wholly rational or irrational: our aim is to make them realistic.
An environment: the population exists within a virtual world that provides the set of external factors that influence behaviour. If the agent rules are based on the location of stores, the prevalence of brand advertising in media, or the weather, then our simulation will incorporate environmental components to reflect them. The environment also provides the means for agents to influence each other.
Population simulations provide a single place in which all the existing qualitative and quantitative data about the population can be knitted together. The use of an explicit representation of an individual and their decision making process within the simulation makes it easier to fill the inevitable gaps in knowledge using expert opinion or informed assumptions. The models can be continually improved with new and updated data and insights so that the simulation always represents the best and latest understanding of population behaviour.
Whilst based on historical data population, simulations are explicitly forward looking through the focus on identifying the rules that govern individual behaviour. This means that our forecasting capability is not constrained by historical data. The simulated population will attempt to make sense of a new situation even if they have never seen it before whether that is a new product, the impact of a long-term change in attitudes, or a market shock like a product recall.
The simulations reconcile our understanding of what drives individual behaviour with the dynamics of the population as a whole, connecting the micro and macro through the social interactions of individuals – the meso. This allows us to disaggregate the origins of population behaviour and so identify the segments or decision points on which we should focus our efforts in order to change that behaviour.