In the context of the CROSSROAD project, yesterday in Rome, I was invited by Stefano Armenia to attend the meeting of the System Dynamics Italian Chapter, on applications of System Dynamics to public policy. The meeting included presentations on real-life application of System Dynamics tools and methods to public policy, and allowed me to better understand the status and future opportunities.
Basically, SD is used to anticipate the impacts of decision; to elaborate scenarios of future impact which take into account a large amount of interrelated variables. First, it builds complex causal models (the typical causal loops). Then it adds feedback and interaction between them (the dynamic element). Then it runs software that allows to simulate the scenarios of different kinds of solicitation into the system: not only taking into account the causal relationships, but also the dynamic feedback mechanisms that oftern spur unintended consequences. It is, in summary, a way to capture complexity and wicked problems, in order to have a better view of long-term and unintended effects, and simulate the impact of different actions. Here’s a typical diagram from wikipedia.

In particular, this triggered my thinking about the CROSSROAD model. My vision of future policy-making focusses on tools that support the integration of three, traditionally alternative, features of policy-making:
– evidence-based (traditionally, through experts’ input)
– timely (traditionally, through hyerarchical decision)
– participated (traditionally, through lengthy consultation)
Now, I realize that SD and the other simulation tools add a fourth feature: the long-term thinking and anticipation of future events. Too often public policies are based on short-term thinking – typically the attention span of the media, or at best the electoral cycle. Too often they fail to take into account the many implications of the decisions. This is obviously very important to address issues such as Climate Change.
So this is a key application field: anticipating the impact of policy decisions. Take the example of the financial crisis: first the crisis happened, then government tried to intervene, then government suffer from financial exposure – with all the societal implications of this. None of these events was expected.
This can be done, traditionally, by econometric tools. But they are “too linear” – they fail to capture complexity and non linear phenomena, and thereby are unsuitable for black swan events.
This can be done better by SD tools, which are able to account for the dynamic feedback mechanisms between actors and between events.
Social simulation tools, and agent-based modelling, go one step further. They not only take into account the dynamic effects, but they are open to unexpected causal relationships and interactions between agents. The model itself is dynamic.
However, the overall questions is: how can decision support tools help us dealing with increasing improbable events?
How can we deal with black swan events and wicked problems? Isn’t it a contradiction to try to simulate and anticipate impact when we recognize it’s impredictable? Shouldn’t we just rely more on human judgement?
My impression is that simulation tools are important, on one side, because they are able to capture and simplify a wider set of interactions than traditional econometrics. They are able to structure complexity and provide a more manageable, and more comprehensive, view of future impacts.
But models always carry the risk of excessive reductionism. This is why we need modelling tools that are able to fully capture human expertise and to augment is.
We need to offer modelling tools that are usable directly by thematic experts – not by the methodology/technology experts.
We need collaborative modelling tools, open to the wider set of human intelligence. And we need to go beyond the notion of open data: we need interoperability and open models so that we maximize the effort of analysts and stop re-building the wheel.
Furthermore, these modellng tools should be designed in a way that is usable and open to the wider public. Citizens should be able to visualize, maybe in virtual reality, the impact of the different actions. We would need a policy modeling tool is used as a debating tool by stakeholders, in public. Like, but with underlying models: what could happen if Greece goes out of the EURO? You would see policy-makers interrogating the tool with different possible options, and possibly citizens too.
In their daily life, we could envisage displays for citizens that show the impact of different decisions – just as people are studying display technologies for changing the behaviour of citizens in energey efficiency.
And they should be able, should their expertise allow it, to interact not only with the reporting, but with the models themselves.

In other words, my impression is that we need to futher develop modelling and simulation tools in order to let the widest set of intelligence, and especially the thematic experts:
– take part in building the models;
– carry out the analysis;
– play with the reports dynamically.

This is obviously an initial and superficial reflection, but the key point is that we need to apply the “augmented” metaphor to policy modelling.

Other questions to be addressed in future posts are:
SD and simulation is far from new: it was created in the 1950s. Why now?
What is specific about simulation in governance, and different from commercial applications?