Agent-based social simulation

Bruce Edmonds, Manchester Metropolitan University.

Video: Bruce Edmonds

Agent-based modelling by Bruce Edmonds

Video: Bruce Edmonds

What is agent-based social simulation? by Bruce Edmonds

Agent-based simulation allows the explicit representation and exploration of the complex relationship between individual behaviour and society – the Micro-Macro link. It does it by representing the states and actions of each relevant social actor within a complex computer simulation.

In this technique social actors (people, firms, parts of the environment, etc.) are each represented by separate entities (called ‘agents’) within the simulation. The interaction between the social actors are modelled as messages between the agents. The entities are given behavioural schemata in the form of sets of inter-related rules and each agent has its own individual characteristics (e.g. memory, habits etc.). When the whole simulation is set going all these rules are repeatedly evaluated in parallel, so the effects of each rule will depend upon the past outcomes of rules, resulting in a complex sequence of interactions between the agents.

The technique allows for a precise but relatively rich representation of the complex interaction of social actors over time. The results in terms of the outcomes at both individual and societal levels are then inspectable. The approach provides concrete examples of the results of all this interaction in terms of specific simulation traces. It thus establishes the plausibility of micro-macro explanations and allows the “in vitro” exploration of the possibilities. 

This technique can be used in a very detailed and specific manner, enabling the coherent integration of various kinds of evidence (e.g. narrative, time-series, survey, SNA) within a sort-of dynamic computational description. This use contrasts markedly with the reductive modelling of Neo-classical economics, allowing the simultaneous emergence of unpredictable social phenomena and the top-down constraint of individual action from norms and institutions. However their very complexity can make the simulations themselves difficult to understand completely and their level of detail can make validation challenging.

Experts/centres in the UK

Some relevant projects

Further reading

Resources on the web

PDF slides