At its simplest, data visualizations can help Labs participants gain an understanding of the current state of the system. As Frances Westley says, “Tools for producing infographics that clearly show the relationships between different data are becoming much more widely available [such as those by Hans Rosling]”. Examples of this can also be seen at sites such as Informationisbeautiful.net and VisualComplexity.com. Even before common understanding of a system, it can be useful to visualize and gain consensus on the desired outcomes of system interventions. Sometimes this is as simple as surfacing assumptions and driving a discussion around these.
As an example, the model below illustrates the impact of various development choices on a local community’s goals. Communities are often faced with the challenge of evaluating the impacts of development opportunities. More than that, communities must decide what is important (creating jobs, earning revenue, plugging leaks in the local economy) and how to balance these competing demands. This model (built using the SAP Crystal Dashboard tool) provides a simple illustration of a model to do just that. In this scenario, a community has come up with three investment options:
- Start a Green building company
- Create a landscaping company
- Invest in job skills training
In this example, success will be measured by the number of jobs created, the amount of annual revenue and the number of community members receiving job training. By changing assumptions about how much revenue and how many jobs are generated by each option, the community can immediately see the impact on goals. More importantly, the model surfaces the goals for discussion. Community members are forced to consider the relative importance of each goal, the balance between the goals and also the assumptions behind the enterprise options. In this example, there are linear interactions between the variables and there is little connection between the various options. These types of models can be very useful when making decisions between alternatives that have been generated at an earlier stage. While useful for discussion purposes, if we wish to model complex adaptive systems, we must look to more complex data visualization techniques.
Modelling Complex Systems
As Westley et al. state, “Simulation and visualization is an area that has tremendous potential for helping people to understand complex systems”. They go on to say that while “mapping and prototyping is already widely used in design processes…future work will make better models that are easier for participants to manipulate, and will more deeply embed compelling visualizations into the toolset to support Change Labs” (emphasis added). The use of simulation and visualization has a rich history in this context. Forrester and colleagues at MIT have been building systems dynamics models (Richardson, 2011 cited in Westley et al., 2012). The use of “feedback supporting a person to change behaviour” came from “WWI and WWII control theory models that included feedback to help planes fly better” (Lewis, 1992 cited in Westley et al., 2012). The emergence of parametric modeling provided a more flexible interface getting closer to modeling a complex system. “In a good parametric model, changing just a few variables can transform the whole system” (Woodbury, 2010 cited in Westley et al. 2012). However, a key limitation of parametric modeling (or any model for that matter) is that the variables, relationships and system can only be changed in ways that have been anticipated by the model designer and the limits of the technology used for the model.
Westley et al. cite the examples of Conservation Breeding Specialists Group (CBSG) that “…developed a tool that let policy makers make decisions in simulation and understand the effect those decisions could have on particular species. These proved remarkably effective for increasing decision makers’ understanding and as a tool to support decision-making” (Lindenmayer, et al., 2000 cited in Westley et al., 2012) and John Robinson and Jonathon Salter at UBC (Tools for Modeling, Visualization and Community Engagement, 2011) who “developed visualization software to be used with members of the public and decision makers to understand the implications of their own action beliefs and values” (cited in Westley et al., 2012)
One of the reasons why the tools and models have been challenging to develop is that in typical design processes, the result is often a “thing” – easy to see, build, and have control over. Prototypes can easily be built and models tend to be linear and predictable. Designing at a system level is harder to prototype and can be very expensive, although not necessarily so. The variables and relationships are complex, results are nonlinear and the system is inherently difficult – if not impossible – to model completely. This is a core feature of the irreducible complexity of complex adaptive systems (F. Westley, personal communication, September 30, 2012). One of the characteristics of complex adaptive systems is the unpredictability and sensitivity to system shocks. In Anti Fragile, Taleb, argues that it is impossible to accurately predict the likelihood of system shocks and that the best a model can do is assess the fragility of a system (2013).
Given the complexity and expense of building prototypes, these models are often not really treated as an experiment to learn from what worked and didn’t but rather as a solution to be implemented. The power of good visualizations is for both seeing systems and experimenting with systems. Models can be built that allow participants of a lab to “test” out solutions. The concepts of rapid prototyping are very useful here as per Harrelson (2010). Harrelson outlines three principles for effective prototypes:
– Fast: allowing for rapid iteration (and feedback)
– Disposable: enough to express the idea to be communicated, and no more
– Focused: selecting the most important things to test – such as significant “unknowns” or complex elements (cited in Young, 2010)
These models are not necessarily scientifically valid but still can be incredibly informative to play with as simulations. The intricacies of complex adaptive systems can be hard to hold in ones head so interactive models can be useful to ask “if we do this, what is the impact on that?”, “what is connected?”, “what goes up?”, “what goes down?”. Even if the answers to these questions are contentious, the discussion around these variables and their relationships can be hugely valuable in forming a common understanding of the current system. I will describe this in more detail below.
The concept of Social Innovation Labs is to engage the whole system. However, the initial focus of many Labs is of policy makers as participants. While this has limitations (as will be discussed later), it does allow us to make some predictions about what might be useful areas for simulation and visualization to play a role. Simulation can be useful to illustrate political horse trading by offering a sense of the pushback that might occur given a certain policy intervention. Simulation can also ground decision makers in risk and allow an understanding of the variability (M. Tovey, personal communication, November 30, 2012). In essence, the impacts become intuitive. This can allow decision makers to come to agreement on what the policy actually is (variables, assumptions). Tools using the concepts of gamification are especially useful in this context. Examples include the Treaty game developed by the Hul’qumi’num Treaty Group at University of Victoria and Democracy 2 This interactive policy game is being modified by Tovey at the University of Waterloo for use in policy simulation
Finally, by making the system visible, it provides the opportunity to “point at stuff” – to go from a top-level overview to detail. (M. Tovey, personal communication, November 30, 2012). If designed effectively, the visualizations can change the metaphor from drill-down to pan-and-zoom (E. Tufte, personal communication, July 23, 2012). Edward Tufte, author of numerous books on data visualization, talks about the traditional drill-down model as being analogous to a Table of Contents where readers (users of the data) can flip to a page of interest. The challenge is that it is difficult for humans to remember the detail on the “page before” and keep in mind the overall context. By changing the metaphor to pan and zoom, it becomes easier for viewers of the visualization to spot anomalies or patterns and zoom in for more detail without losing sight of the relationship of the detail to the whole.
Next time we will talk more about different methods of visualizing systems including the Basins of Attraction model, system maps and simulations.