��������������������������

putational models allows for continued refinement , repurposing , and data collection as new insights are gained . Although these simulation models may be used for predictive purposes , that is not their only use . Such models , even if stylized , can be used to provide insight into hypothetical policy scenarios ( Epstein , 2008 ). What follows is a basic description of a stylized restaurant inspection simulation model . It has been used to compare and evaluate two relevant policy scenarios .

Model Overview , Design Concepts , and Details

Details ( ODD ) framework has been suggested as a foundation for describing ABMs in a consistent manner to facilitate re-implementation ( Railsback & Grimm 2012 ). The ODD protocol will be used in this section to describe the simulation model and environment .

The restaurant inspection model was implemented in the NetLogo 5.0.1 software . As a purposely designed agent-based software package , NetLogo supports three kinds of agents . These are referred to as turtles , patches , and links . Turtles are mobile agents and are used here to represent consumers and inspectors . Patches are locations ( on a grid ) as defined in the NetLogo computational environment . These have been used here to represent restaurants . Link agents were not necessary to model this particular problem .

������� : The purpose of the model is to create a stylized food safety inspection system and compare two relevant policy scenarios . The first scenario incorporates elements of the current inspection system in the Canadian province of Saskatchewan . In this case , restaurants are inspected by local health authorities and are assigned a re-inspection priority on a scale of low , moderate , or high depending on risk factors identified during the inspection procedure . The second scenario to be examined builds on the first by giving consumers access to the re-inspection priority scores , and also incorporates elements of risk aversion in consumer choices . Each scenario is evaluated based on the number of inspected restaurants , number of contaminated restaurants , and number of sick consumers over the course of the model run .

������������������������������������� : The model contains three kinds of agents : consumers , inspectors , and restaurants . Consumers are endowed with state variables . These describe :

In the second scenario , a risk aversion parameter for the consumer is added . The inspector only has a state variable describing their geographic range of operations . Finally , restaurants are defined by a global variable that describes whether or not a given patch is a restaurant , and also possess state variables describing where they are located , whether they are contaminated , their re-in-

5

View this model in the CoMSES Model Library : https :// www . openabm . org / model / 4304 / version / 1 / view

143