DesignWIKI

Fil Salustri's Design Site

COLLECTIONS

2015.01.09 11:13
research:systems_modelling

Systems Modelling

Notes on developing systems models that can help study future consequences of proposed changes.

Also see http://msnbcmedia.msn.com/i/MSNBC/Components/Photo/2009/December/091202/091203-engel-big-9a.jpg as an example of a causal loop diagram.

• Relate causal loop diagrams to “system influence diagrams.”
• Treat TODOs in footnotes.

A generic process for systems modelling

1. Meet with stakeholders/experts to identify what needs modelling, the bounds/scope of the model, the model's significant elements, the nature of the causal relationships between different elements.
2. Create a causal loop diagram1). This is a diagram that contains all the elements described above and the relationships between them. Creating the causal loop diagram can be done on paper or in software2). Here is a classic interactive system influence diagram that is similar in purpose and style to a CLD. See gallery of CLDs below.
3. Vet all aspects of CLD with stakeholders and experts.
4. Create a stocks and flows dynamic model3) based on the causal loop diagram. The stocks and flows model needs to be created using system dynamics software. Possible software includes Vensim (free download of the academic version). See gallery of SFDs below.
5. Once these stock flow models are created in system dynamics software, the relationships and values of variables can be added to the model in mathematical form. This creates the dynamic mathematical model that can be run as a simulation for the designers to see the outcomes of different changes over time. Some simple examples are shown at http://en.wikipedia.org/wiki/System_dynamics (these are not particularly good as they go past so fast - but they are the easiest to find!)
6. Then, using historical data, check the dynamic behaviour of the model with real world changes in the past and tune the model as needed. This is important to validate the dynamic behaviour; if it can “predict” the current state from known historical data, then it's more likely to predict the future state from current data.
7. The use the system dynamic model to show how the consequences of your design decisions play out in time in the future.

Other notes

All outcomes vary over time. It contrasts with common assumptions that the consequences of a design (e.g. a poster) are fixed and simple (e.g. the number of people attending a concert) and instead reveals (depending on the model) the implications over time on all sorts of different planes (e.g. changes over time in people's attitudes to particular types of music or genres of poster design, changes in social interactions between people, changes in amounts of disposable income people might consider allocating to concert attendance etc).

Terry Love also writes on 24 Sep 2013:

The causal loop diagramming is the first stage. It is expected that one would NOT understand the behaviour of all of what is represented in a causal loop diagram In the case of the US Military model, lots of stakeholders and experts contribute their on knowledge about which things are important and which things cause what. That stakeholder and expert knowledge comes together in a single causal loop diagram that everyone can check includes their knowledge and how it fits with the knowledge of the situation of others. This allows this assembly part of the process to be done well within the capabilities of humans.

The whole causal loop diagram , however, is beyond what any individual can understand. This is exactly as is expected and intended. If the causal loop model was fully understandable then there would be no need to use system dynamics modelling of the design situation - the design could be done by traditional design methods.

The role of the next stage, stocks and flows modelling, is to create a dynamic model that can be run to simulate how the world will behave over time. That enables designers and stakeholders to think how they might intervene by changing things and then test out their ideas.

The reasons for using these kinds of approaches are described by Forrester. The challenge of the COIN model is to create a realistic stocks and flows model from the causal loop diagram and then test strategies to see their consequences. It is at that stage that a different kind of more informed partial understanding becomes possible.

Interestingly, looking over the COIN causal loop diagram, its clear it is created from an American military strategists organisationally-focused perspective. I'd guess that a causal loop model created by different groups in Afghanistan would be different. For example, “terrain,” particularly mountainous terrain of deep valleys, is something that shapes many aspects of people's lives, group relationships, culture, priorities, family structures, religion, perceptions of governance etc. and yet in the COIN causal loop diagram it is relatively ignored (in brown about halfway down right hand side) and not linked to much. It would be interesting to create a composite causal loop diagram involving all stakeholders, and then create a systems dynamic model containing the knowledge and purposes of all sides to show how it would play out in terms of the different interventions. It would not be possible to understand the causal loop diagram. It would however be possible to see how the outcomes were likely to play out for different strategies on each side.