surrey logo

logo EPSRC

d

INFOrM - Incident Food Modelling:

an agent-based model

Introduction

INFOrM is an agent-based model of the UK fresh beef mince supply chain. It is designed to help us  better understand what happens at the time of a food incident, what policies might reduce the likelihood that an incident develops into a scare, and how detrimental impacts of any resulting food scare may be minimised. The model is based on the 2013 Horsemeat Scandal and its subsequent escalation from a food incident into a food scare.

INFOrM studies the movement of fresh mince beef products along the supply chain by simulating the buying and selling practises of the main stakeholders in the supply chain (farmers, markets, abattoirs, processors, retailers, butchers and consumers). Whilst the model is running these 'agents' can interact with those immediately above and below them in the supply chain. Each agent buys from a seller if the seller has products to sell, if the products match their specifications and, preferentially, if they have bought from them before. If over time a stakeholder does not trade, it ceases trading and a new stakeholder is created.

Why should I use an agent-based model?

INFOrM is just one example of agent-based modelling, a powerful approach to developing simulations of policy processes which places the heterogeneity of actors, and their interaction with each other and the policy landscape, at the center of the modelling process. You can read more about agent-based modelling here

The model interface:

INFORMscreenshot1

Beef products 'split' multiple times as they are traded along the supply chain. This represents the life cycle of the product from its starting point as a living animal to a 4-pack of burgers. INFOrM uses food waste (in kg) and the number of businesses ceasing trading as output indicators. A food incident can be initiated either endogenously within the model or by the user. The magnitude of the food scare is controlled by a range of user-set variables. A second set of user-set variables control the policy-based scenarios that are implemented.

Using INFOrM

These instructions allow you to download and run the model. This demonstration is intended to help you understand what an agent-based model is, how it works, and how the approach may of use in your topic or problem area. If you think an agent-based model of the bio-waste industry is closer to your interests, you may prefer to look at the ePROBE ABM

First, you should download and open the model. To do this you will need to download 'NetLogo', the program in which the model runs. To download NetLogo, go to https://ccl.northwestern.edu/netlogo/ and follow the instructions.

Once you have done this, download the model here (you must use 'right click and save as' with '.nlogo' as the file type), and open it in NetLogo.

Controlling the model

The pale blue/grey boxes in the top left corner are the main boxes for running the model. The green boxes and sliders are inputs for the model. There are two types: Graph variables and Policy Scenarios (which are based on the Elliott Report). The cream boxes are the monitors and graph outputs for the model. For each run:

  • Change variables as required for the scenario you wish to run. For your first run leave all settings as they are. Information about each of these variables is provided in the Appendix tab below.

  • Click Setup – this initialises the variables. We suggest you leave the variables as provided for the first run. NB: YOU MUST PRESS SETUP BEFORE EACH RUN OF THE MODEL TO RE-INITIALISE ALL OF THE PARAMETERS AND PLOTS. You should do this just before you run the model, and after you have altered any parameters you are exploring.

  • Click Go – this will run the model for 100 days...OR...Click Step – this will run the model for one time step (i.e. one day)

  • Wait for 50 days – this is to allow the model to equilibrate, days are displayed in the cream box to the right of the Food Incident box.

  • Click Food Incident when you would like the food scare to occur. A food scare will occur at 70 days automatically.

  • To stop the model: re-click Go. The model will stop automatically at 100 days.

  • Observe the various ouput plots.
  • If you wish to return the model to the default settings press: reset

Comparing runs

To compare different outputs we suggest you move a copy of the ‘Amount of Waste’ and ‘Stakeholders who have Ceased Trading’ graphs into a word document. To copy the graphs:

  • Hover over the graph you wish to copy

  • Right click and select ‘copy image’

  • Go to the word document

  • Right click in the box into which you wish to add the graph and select ‘paste’.

Do the graphs look the same if you run the model twice without changing the Graph variables or the Policy Scenarios? You may wish to record outputs of interest in the table below to help with your comparisons. Remember to hit 'Setup' before you run the model, but after you have altered input variables.

Final Amounts

Run 1

Run 2

 

 

 

 

 

 

 

 

 

 

 

 

 What difference does changing the media_magnitude settings make?

Final Amounts

High Media

Medium Media

Low Media

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

We are now ready to test the effect of various policies. The Appendix provides details about the variables that are used in the Policy Scenarios.

What differences do changing settings in one of the scenarios and keeping the rest as “none” make? Do certain scenarios seem to have more of an effect than others?

We have suggested the scenarios and values you try.

Exploring Voluntary Info Provision Info_vol: 100% Info_vol: 20%
Final Amounts Run 1 Run 2 Run 3 Run 1 Run 2 Run 3
             
             
             
             

 

Final Amounts Level of strategy: High Level of strategy: Low
Run 1 Run 2 Run 3 Run 1 Run 2 Run 3
             
             
             
             

What difference does it make running two scenarios at the same time? Which has the most effect?

We have suggested the scenario you try...

Final Amounts Crime_squad: Off/ LA_checks: Low Crime_squad: On/ LA_checks: Low
Run 1 Run 2 Run 3 Run 1 Run 2 Run 3
             
             
             
             

Appendix

In the model interface there is a series of green boxes in the lower left hand corner where you can change variables to test policy options. Some of these are ‘sliders’ and some are drop down boxes.

Four over-arching policy scenarios are provided.

Audits

  • Check whether audits have been conducted and to what standard; if not then their traceability is reduced

  • Can change the amount of sampling

  • Local Authority checks that companies are complying. Those who are not complying cannot sell

Media

  • Embargo on press reports for a set period of time. Suggested times: 2, 5, 10 days

Risk and response

  • Crime squad: if the stakeholder’s traceability becomes too low then they will be penalised

  • An “Information database” is in use - two levels: compulsory and voluntary. This affects the stakeholders’ traceability

  • Each stakeholder has a level of how they will respond to a scare - this affects their ability to sell in a scare.

Product removal

  • All products removed from sale until they are proven to not be contaminated

 

What are the Graph Variables?

These 'choosers' alter what information is produced in the output graphs on the right. Wastage denotes whether a cumulative or daily total is displayed, and the two stakeholder choices denote which types of stakeholder are shown in the ceased trading plot ('turtles' shows all types).

 

What is "None, Low, Medium, High, High/Low"?

None - this means that no policy scenario has been implemented and therefore the model randomly gives the stakeholders one of the variable options.

High, low, medium – all stakeholders will be given the level selected

High/Low – half the stakeholders are given high and the other half are given low

 

Other variables

Media

Media cannot be turned off or changed to “none”

Media_length

short, medium, long

The user can change the length of time that the media reports for over a food scare, short is 11 days, medium is 17 days and long is 33 days

Media_magnitude

low, medium, high

The user can change the magnitude of the scare with medium being double the magnitude of low, and high being quadruple the magnitude of low

Media_embargo

2, 5, 10 days

The user can change the number of days the media is not allowed to report for.

Audits

  

Assessments

None, high, medium, low, high/low

The user can change the effect of having different levels of assessment, sampling quantity and local authority checks

Sampling_Quantity

None, high, medium, low, high/low

LA_checks

None, high, low, high/low

Risk and Response

  

Level_of_strategy

None, high, medium, low, high/low

The user can change the effect of having different levels of response strategy. Level_of_strategy refers to how prepared (ie ready) the company is to deal with a food incident. Info_comp refers to the amount of information shared compulsorily.info_vol refers to the percentage of information given voluntarily.

Info_comp

None, high, low, high/low

Info_vol

none, 0%, 20%, 40%, 60%, 80%, 100%

Product Removal

 

Product_removal

none, remove

 

Contact: Liz York (This email address is being protected from spambots. You need JavaScript enabled to view it.)

 

Creative Commons License
INFOrM is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

 

 


Related Content

Your Questions: UnderstandingWhat if?Evaluation

Our Approach: Understand the system

Concepts: EmergencePath DependenceSelf-organisationTipping PointsNetworks

Case study: Food Scares

Go to top