Presentations‎ > ‎

London Knowledge Lab - April 2011.

Constructing computer models by composing with micro-behaviours: a new way to learn about complex systems?
Howard Noble and Ken Kahn, University of Oxford

The slides for the presentation are publicly available.

In what context was the game/simulation/virtual world used?

60 second-year Zoology students at the University of Oxford opted to take a module about infectious diseases. They were divided into two groups and given 2 x 3 hour computer-based practical sessions where they constructed simulations of the spread of a virus through social networks.  

An academic teacher, two demonstrators and three members of the modelling4all project team helped run the session. We used a white-board, overhead projector and each student had access to a networked PC (although many chose to work in pairs).  

What practical issue and/or pedagogic concern prompted its use?

  • Gain an appreciation that computer models are not black boxes and gain a basic understanding of how models are constructed.
  • Compare and contrast building the Susceptible-Infected-Recovered (SIR) model when built using the R statistics tools and an agent-based modelling (ABM) tool (
  • Use the ABM approach to model the spread of a virus through different social networks where the number of connections between people is constant, or based on a normal or power law distribution.
  • Develop a healthy scepticism of the modelling process i.e. a modelling literacy.

What did you do?

In the first 3 hour session:
  1. A teacher gave a general introduction to agent-based modelling (ABM) 
  2. A teacher led a class discussion to design a predator-prey system on a white board (pseudocode)
  3. Students used the modelling4all software to build and experiment with a predator-prey simulation 
  4. A teacher led a discussion about the similarities and differences between the approach previously taken to build the SIR model with the R software (aggregate mathematical approach) and the modelling4all software (agent-based approach).
In the second 3 hour session:
  1. A teacher briefly introduced the session
  2. Students worked alone or in pairs to: 
    1. Build the SIR model by following a guide
    2. Experiment with the model to gather data
    3. Use a plot of the data to answer some questions. 
  3. A teacher then used a plot of data generated by the model to illustrate a few key points about the model.
A few days after the session:
  1. The students are sent each other's answers to the class questions, and model answers written by the demonstrators.  

What, if anything, changed for the better as a result of its introduction?

Previously the session had not existed - the SIR model had been taught using the R software alone. The introduction of the ABM approach is welcomed by teachers and students because it is generally seen as more intuitive, less mathematical, makes it easier for students to experiment with variations of the standard SIR model e.g. introduce different types of social network. 

The practical was first advertised within the department as a modelling exercise. Only a handful of students signed up, most likely because biology students have limited experience of using computers in their learning and prefer to steer away from the unknown. When the course title was changed be to a general session about epidemics 60 students signed up. The feedback from the session by students was very good which hopefully means they are less averse to computer modelling now.

A number of students were very inquisitive as to how the modelling4all software worked, how it related to the NetLogo software, and what else it could be used for. This has resulted in a few researchers building ABM into dissertations and research grants.

What issues were associated with the change?

  • Some students found thinking about modelling at the individual-level and programmatically troublesome. This was probably due to their recent previous sessions where they used the R software to build a population-level model but it could also reflect a deeper issue. 
  • The sessions tend to bring up methodological questions about the 'nature' of modelling that are very difficult to answer e.g. the precise similarities between ABM and the techniques most students are conversant with where mathematical approaches are used (e.g., differential equations).
  • There was of course the inevitable issue of technical support (up to data Java on machines, reliable internet connection, modern browser etc). 

What evidence do you have that substantiates these claims about benefits and issues?

  • 2011 was the 4th year we've run this session and the number of students that sign themselves up is increasing 
  • We run similar sessions in the business school and have just started working with an academic who wants to use ABM in her politics course
  • The answers to the formative questions all students do at the end of the practical session how a good level of understanding
  • We recently started to run introductory courses for anyone in the University; 150+ people have so-far attended and  the feedback has been good. The most recent cohort has asked for a more advanced taught session (scheduled for early June 2011)
  • The ABM courses have resulted in several academics starting to use ABM in their research (anthropology - explaining religions project)
  • We have participant feedback form data from a wide range of sessions (Zoologoy undergrads, MBA students, Business MSc, Introduction to ABM) which tends to be positive
  • We used the software as part of a two week event organised by the Royal Society last year. Our stand was about infectious diseases and 1000+ people of all ages used the software alone or with the help of demonstrators.

What are the implications of this work for others?

Graduates, postdocs and other academics seem to be increasingly interested in using ABM in their teaching and research. The software can be used to allow people to quickly appreciate the potential of this approach to computer modelling i.e. a stepping stone to learning how to program in the more conventional sense. Put more abstractly this might suggest an increased awareness of the value of computer modelling, as Epstein outlines succinctly in his 16 reasons other than prediction to build models:
  1. Explain (very distinct from predict)
  2. Guide data collection
  3. Illuminate core dynamics
  4. Suggest dynamical analogies
  5. Discover new questions
  6. Promote a scientific habit of mind
  7. Bound (bracket) outcomes to plausible ranges
  8. Illuminate core uncertainties.
  9. Offer crisis options in near-real time
  10. Demonstrate tradeoffs / suggest efficiencies
  11. Challenge the robustness of prevailing theory through perturbations
  12. Expose prevailing wisdom as incompatible with available data
  13. Train practitioners
  14. Discipline the policy dialogue
  15. Educate the general public
  16. Reveal the apparently simple (complex) to be complex (simple)
(List taken from Epstein 2008 paper, Why Model?: