1. Patrik Wikström | QUT DMRC | 2015
QUT Digital Media Research Centre
http://www.qut.edu.au/research/dmrc
Patrik Wikström | 26 June 2015
Agent-based modelling
and simulation
2. Patrik Wikström | QUT DMRC | 2015
Many phenomena in comm & media studies can be
conceptualised as “complex adaptive systems” (CAS)
• Large number of (locally) interacting elements.
• Any element is affected by and affects several other elements.
• The interactions are non-linear: small changes in can cause large
effects.
• CAS are dynamic and have a history. They evolve and their past is
co-responsible for their present behaviour.
• It may be difficult or impossible to define system boundaries.
• CAS operate under far from equilibrium conditions.
• Emergence: micro level actions generate macro level
patterns.(Micro motives & macro behaviour)
(e.g.Miller&Page2007)
3. Patrik Wikström | QUT DMRC | 2015
It is challenging to build a theory that is
able to capture such complexities
4. Patrik Wikström | QUT DMRC | 2015
“The paper with the largest circulation in a market
has financial and economic advantages that enable
it to increase advertising and circulation sales by
attracting customers from the smaller paper. As the
leading paper attracts more circulation, it attracts
more advertising, which in turn attracts more
circulation, trapping the secondary paper in a
circulation spiral that ultimately leads to its
demise.”
“Traditional” modelling approaches are simply not very useful:
Verbal models
5. Patrik Wikström | QUT DMRC | 2015
Signifier
sound image
Signified
concept
“Traditional” modelling approaches are simply not very useful:
Verbal models
6. Patrik Wikström | QUT DMRC | 2015
“Traditional” modelling approaches are simply not very useful:
Statistical (e.g. regression) models
7. Patrik Wikström | QUT DMRC | 2015
Interactions between feral cats,
foxes, native carnivores, and
rabbits in Australia.
System of differential equations
“Traditional” modelling approaches are simply not very useful:
Mathematical models
8. Patrik Wikström | QUT DMRC | 2015
Computational modelling, however,
seems like a promising approach (1)
• In a computational model, concepts,
assumptions, logic, and propositions are
represented by a computer program.
• The model allows you to simulate the passing
of time and observe how constructs that are
included in the model change over time. This
enables the researcher to analyse complex
dynamic processes that non-computational
modelling approaches are unable to capture.
9. Patrik Wikström | QUT DMRC | 2015
Computational modelling, however,
seems like a promising approach (2)
• The simulations generate data that can be validated
against real-world data. If the model is able to replicate
real-world processes, it is reasonable to argue that the
assumptions, propositions and logic that are
underpinning the computational model is a plausible
explanation of the observed real-world processes.
• The validated model can then be used to make
structured experiments and generate “what-if”
scenarios in order to make contributions to theory.
10. Patrik Wikström | QUT DMRC | 2015
There are different types of computational modelling;
one is Agent-Based Modelling (ABM)
11. Patrik Wikström | QUT DMRC | 2015
Agent-based simulation models are able to
capture the peculiarities of CAS fairly well
• Heterogeneous entities (“agents”) interact with the
environment and with other agents.
• Agents have perception, a set of behaviours, memory &
cognition, and follow certain rules or policies.
• Agents interact in a space that can be a representation of a
physical space, but doesn’t have to be.
• Agents can be modelled as nodes in a network.
• Simple rules on micro (“agent”) level generate complex
patterns on macro (“population”) level.
• Feedback between micro and macro scales.
(Emergence & immergence)
12. Patrik Wikström | QUT DMRC | 2015
NOLAWhite:bluedots;AfricanAmerican:greendots;Asian:red;Latino:orange;allothers:brown
Agent-based simulation modelling is still in its infancy in our
field but is increasingly accepted in social sciences such as
economics, sociology, pol sci and anthropology.
13. Patrik Wikström | QUT DMRC | 2015
The process for simulation supported
theory development is fairly conventional
“theory as process; that is, theory as an ever-developing entity, not as
a perfected product”
(Glaser & Strauss, 1999 [1967]: 32)
1. Model development
– Build a computational model that takes as its input a set of constructs,
propositions, logic, assumptions about locally interaction agents.
2. Data collection & analysis
3. Model validation and simulation experiments
– If the model is able to replicate macro-level behaviour that can be
observed in the real world – then we might argue that the theory
behind the model is a plausible explanation to the observed
behaviour.
14. Patrik Wikström | QUT DMRC | 2015
Useful links
• Download NetLogo:
http://ccl.northwestern.edu/netlogo/5.2.0/
• NetLogo models discussed during the workshop:
https://github.com/qut-dmrc/simple-models
• This presentation:
http://www.slideshare.net/patrik/abm-intro-talk
• Useful papers/theses/etc.:
https://paperpile.com/shared/ypQIj8