Modelling the spread of viruses: Agent-based modelling


Academic researchers create agent-based models (ABMs) to understand traffic flows, wealth distribution, foraging in ant colonies, public policy and many other systems across the physical and social sciences. As this comparatively new method of enquiry Photograph of Robert Belshawmatures researchers are starting to agree on how to document the modelling process for peer-review (e.g. by using the ODD protocol). This is helping researchers to justify investing time in learning how to create ABMs as it is increasingly seen as work that can contribute to ones publication record.

Dr. Robert Belshaw is keen that his students understand the importance of ABM in Biological Sciences. Robert designed 2 x 3 hour practical sessions for his Infectious Disease Control course. Students build and analyse an ABM where they can measure the effect of different vaccination strategies on the spread of a virus through human populations with different social network structures.

The Challenge

The beauty of the ABM practical is that it is scheduled to take place after students have used the R software to construct a mathematical model of infection. The students are encouraged to compare and contrast the two approaches, particularly with respect to the ease at which it is possible to build agent heterogeneity into ABMs. In the ABM session students observe the spread of a virus within a population where people have the same number of connections and where the number of connections varies according to the power law. Most students agree that ABM is a useful tool for constructing and testing ideas that would be difficult using mathematical techniques (ordinary differential equations and statistical methods).

The Innovation

To run this session Robert used the Behaviour Composer (a web-based tool for constructing NetLogo models), developed within the LTG Oxford and a new WebLearn tool. Students follow a step-by-step guide where they compose a model by assembling blocks of NetLogo code called micro-behaviours. At each step the guide spells out what they are doing in detail so that they can understand every assumption that is built into the model. After the students have become accustomed to composing their model they are encouraged to build alternative models to explore new ideas. They do this by searching through a library of micro-behaviours and customising micro-behaviours before they add them to their model.

ABM Screenshot
To help the students appreciate the importance of systematically analysing the model they are asked to run a number of experiments with the model. To do this they run the model with specific parameters set and record their findings in a Google Docs form. The spreadsheet associated with the form is structured so that graphs of the class data are sent back to their class WebLearn site. The students then use these graphics to answer a set of formative questions (also in WebLearn).

ABM Screenshot 2

All student answers are automatically aggregated together so that they can learn from each other. Robert then sends back a set of ideal answers by updating the WebLearn site wiki page.

Robert has used WebLearn as a hub for bringing together a set of tools that the students can use together to learn about ABM. Now in its fifth year, despite the apparent complicated setup the session runs with minimal administrative work, which allows the teachers to focus on helping the students understand the key conceptual ideas underpinning viral epidemics and ABM.

The student feedback is very positive and suggestions for improving the session centre mostly on extending the scope of the modelling exercise to include new disease dimensions – something which is made fairly easy by using the Behaviour Composer.

Top tips for success

1. If you’re wondering how to use ABM in your teaching ask for help from the Learning Technologies Group. We have 5+ years of hands on experience at running ABM sessions at the University.
2. The Behaviour Composer works with the NetLogo software (free and very easy to install). If you can download NetLogo to the PCs the students will be able to perform more rigorous data analysis using the BehaviourSpace.
3. Ask your local IT representative to install and enable Java within your browser to use the Behaviour Composer.
4. We run introductory and advanced ABM courses at OUCS.

Posted in Mathematical, Physical and Life Sciences | Tagged , | Leave a comment

Leave a Reply