Policy and Complex Systems
Agent-based modelling ( ABM ) is similar to SG , with the difference that the people in serious games are represented as artificial agents in ABM . ABM covers many of the benefits of SG because both tools deal with populations of individuals . Since ABM is aimed at existing systems ( Moss 2002 ), and is descriptive rather than predictive ( Lempert 2002 ), it is instrumental for comparing and illustrating the before-and-after situations of a policy implementation . In addition , ABM also provides valuable insights into the emergent outcomes of policy implementation which are the result of individual behaviors and reactions . In fact , ABM combines the benefits of the aforementioned tools . However , there are still areas for improvement .
The issues related to emergence in simulations are ( 1 ) the type of emergence : are they physical ( e . g ., traffic patterns )? or social ( e . g ., punishment for stealing )?, and ( 2 ) the detection , analysis , and possibly control of emergence . In current ABM research , physical emergence is extensively addressed in simulations especially when using visual tools such as Netlogo . It is , however , more difficult to address social emergence because it is not always visually recognizable in a simulation .
One other drawback of ABM is that since the systems are simulated from bottom-up , there is no straightforward method to simulate the social or technical environment of the system , or to define the boundaries . This also makes it difficult to include the technical , economical , societal , and political aspects of policies into simulations . Currently , these factors are either not considered in the simulations or they are modeled as part of the agents . Modeling social structures within agents is not realistic because agents and structures are interrelated but separate concepts . The primary consequence of simulating the combination of the two as one entity is that we would not be able to study the influence of social structures on individual behavior and the system as a whole . Furthermore , social structures are also influenced by individuals . If they are modeled within agents , it is not possible to model global changes in these structures and observe how they evolve and diminish , and how new structures emerge . In current agent-based models , it is difficult to explicitly display and present policies because of the inability to model social structures . Therefore , being able to model policies as a purposive design of social structure also facilitates their presentations . final drawback of ABM for policy analysis is related to participatory model development . Although simulation results can be communicated to problem owners to facilitate participatory decision making , building collaborative agent-based models is not a common process .
5 . Enhancing ABM for policy analysis
Although ABM addresses various requirements for policy analysis , there are still areas for further extending its applicability in this area . In this section , we discuss how ABM can be improved at each of the requirement levels of Section 3 .
Enhancements for Problem Definition in Policy Analysis
While there are visual tools and techniques for ABM that facilitate communication with domain experts by showing results of simulation runs , consultation with stakeholders for problem clarification is not a built-in facility for ABM . There-
70