IBM approached the city of Portland in late 2009 to partner on creating such software, facilitating sessions with more than 75 Portland-area experts. Ten years' worth of historical data was collected to provide an interactive visual model of Portland as an interconnected system. This model helped develop the Portland Plan, setting the city's policy priorities and objectives for the next 25 years.
Last week, IBM introduced the model -- System Dynamics for Smarter Cities -- to municipalities for their planning efforts. I spoke with Portland's Chief Planner Joe Zehnder about the software development process and how the city used the software's projections in this condensed, edited transcript.
How did this city-developer partnership transpire?
We are pursuing a strategic plan, trying to break through our typical silo way of thinking. You can be an agency that is responsible for transportation, but transportation investments could have a big impact on economic development. They could have big implications for neighborhood development. We were looking at being multi-objective in how we did this. We set priorities around prosperity, education and neighborhood development, looking across the different topics that divide up our thinking. System Dynamics offered an opportunity to look at that, so we sat down with IBM and talked through how the model might be able to [help us] better understand and explore the way that one's actions in a sphere might affect other [spheres], and to help us set priorities.
How does Portland's data shape the software?
We started by looking at the logic of the model and what is related to potential impacts in terms of choice, decision or policy. We built this set of relationships -- this is related to this and something else is related to that -- and [IBM] built these loops or set systems around specific topics. Then we took a step back and said, "OK, what information do we have that could be relevant to that topic, like transit times or income or health in a different part of town?" Do we have that data? Do we have data that we could break down geographically and over time, and then build a data set around that? Which was our best possible set of choices? We were pretty inclusive. Part of what this is about is how to use real-time information, or the vast amount of information out there in a city, in a new way or more intermediate way to improve our understanding of what is going on in terms of performance and change.
How does all this data from the last 10 years help project these future changes?
It doesn't really project; it's not really a forecasting model. It teases out the potential size and direction of impact that one kind of change might have on other factors in the city. If you made an investment [in transportation], what kind of change might it have in terms of air quality, access to employment, income or education? We found that we really can't actually model some of that stuff, but it allowed us to broaden our thinking. The process of thinking through the model was probably as valuable as the end product.
You said the model is really technical. How so?
It's not so much about the end product as it was about the exploration of this tool and the kind of questions we need to pose and answer to even begin to build a tool like this. The funny thing about working with IBM on this is that we are city planners, policy people and community development experts. It is a whole different language when you are talking to modelers. It was interesting to build a bridge across our different languages and logics. But we, in developing the model, pulled together local subject matter experts. IBM facilitated an exercise with them to explore what they thought the most important relationships were to impact human health in a city or economic development, and what the factors that play into that kind of outcome are. And when you think about that and about the potential to slip into a world where everything is related to everything, it can get pretty complex. So if you were trying to build the real ideal model of a real urban system, it is vast and maybe unattainable.
If the software identifies a particularly bad outcome, how would that affect decision-making?
That didn't happen, but if it did, it would be the kind of challenge that we are really after. Part of what we were trying to do in developing the Portland Plan is challenge our assumptions about the condition of the city. So if it would have done that, that would have been the kind of thing that we are looking for. We assumed that this is how it plays out and this logic model we built is telling us that are there are some problems.
So you wouldn't disregard the model's suggestions or predictions -- whether good or bad, you'd take everything into consideration?
Right. It is to challenge assumptions and force you to think differently than you have in the past -- to see connections that you may not have put enough weight on. Given a choice about education, transportation or environmental investments, we want to tease out how they impact each other and which of those are going to support these three priorities we have in the plan.
Do you think you'll continue to use this software to update the Portland Plan?
I don't know. One of the things that I think I enjoyed about the partnership most was that they were learning a ton from us and vice versa. It wasn't one of those situations where someone comes in with your solution and the savings tool. It was more: Is there a tool here worth working on and what can we learn from it? That is a pretty open kind of relationship to have in exploring this. What we produced in our efforts with this is something that is too complicated for us to maintain on an ongoing basis, but there is a lot of promise for tools that are maintainable. We didn't get it there in our efforts, so we both had to move on to finish the Portland Plan. We are going to use it to explore our actions, but I don't think that we are going to maintain it.
