Powered by System Dynamics
Oct 23, 2011
11 minutes read
Photo: Futuro screen capture as presented by Stafford Beer in Cybernetics Part 3 (June 1974)

Photo: Futuro screen capture as presented by Stafford Beer in Cybernetics Part 3 (June 1974)

Back in the early seventies, for a very brief period of time, Chile had access to unprecendented technology to manage its activities. Project Cybersyn had produced an operations room which contained a number of components fed with near-real time information received from all enterprises spread across the long strip of land wedged between the Andes mountain range and the Pacific Ocean. One of the components in that room was Futuro which is shown in the screen capture above. Futuro was an implementation of CHECO, a Dynamo program of the CHIlean ECOnomy based on Jay W. Forrester’s contribution to management science: system dynamics (SD). The display had been implemented with electronic components which depicted the usual symbols found system dynamic models: levels, valves, information flows. The flows were implemented with light-emitting diodes producing marching ants and thus the direction of the information flow.

Dynamo, the software used to model and run CHECO, was a special-purpose compiler for simulating SD models on an IBM 204, 709 or 7090. Its core consisted of approximately 10.000 instructions including parsing, executing and rendering character-based plots. An IBM 709 with its 64KB memory could support a model with 1,500 variables. We have come a long way since Dynamo with software packages such as Vensim, Stella, PowerSim as well as SaaS-based solutions such as InsightMaker or Forio to name a few. All these allow the modeling and execution of SD models in a very comfortable and highly-interactive way. All SD modeling software packages share the ability to deal with systems of differential equations which are solved through numerical methods such as Euler or Runge-Kutta.

The Legacy Of Jay W. Forrester

Jay Wright Forrester introduced in the late fifties a new perspective on systems which focuses on modeling them as continuous processes by means of differential equations. Forrester’s approach rooted in General Systems Theory and Information Theory was adopted by Cybersyn’s scientific director Stafford Beer due to its strong synergies with management cybernetics. One of Forrester’s key contributions was in the field of management science with his 1961 book Industrial Dynamics where he investigated the information-feedback characteristics of a retailer-distributor-factory system. Industrial dynamics shows demand amplification or how the variability of orders received at the factory is greater than the demand variability received by the distributor which in turn is greater than that of the retailer. This would be later referred to as the bullwhip effect by supply-chain engineers and discussed during Beer Game debriefs. The Beer Game is a management simulation of a production/distribution system involving three main characters: a retailer, a wholesaler and the marketing director of a brewery.

In 1968, Forrester ventured out of the corporate realm into urban dynamics. This work lead to the publication of a book by the same title in 1969 in which his team covered four city programs: suburban or governmental jobs for the unemployed, training program to increase the skills of the lowest-income group, federal subsidies to ailing cities, and construction of low- cost housing. The results confirmed how ineffective or harmful these programs had been based on economic health or looking at long-term effects on low-income population. Forrester argued that city programs should deal with the factors affecting population and quality of life (stress from crowding, pollution, hunger and healthcare). The model in urban dynamics goes into understanding the ways population density is controlled and highlights 3 key characteristics of social systems. Below is an excerpt from Forrester’s paper Counterintuitive Behavior of Social Systems.

First, social systems are inhenrently insensitive to most policy changes people choose to in an effort to alter the behaviour of systems. […] True causes may lie far back in time and arise from an entirely different part of the system from when and where the symptoms occur. [..] Second, social systems seem to have a few sensitive influence points through which behaviour can be changed. These high influence points are not where most people expect. […] Third, social systems exhibit a conflict between short-term and long-term consequences of a policy change.

After the city, Forrester and his students took on an even larger system: the World. In July 1970, they organized a two-week conference on world dynamics. An SD model called World2 had been prepared for the conference and presented to the participants which included members of the Club of Rome. The model was structured into 5 sectors: population, pollution, capital investments, agriculture and natural resources. It represented important limits of the world system and how exponential population growth pushes sectors such as agriculture and pollution against the earth’s capacity to absorb its effects.

Limits To Growth And Offsprings

The Club of Rome was already concerned with global interactions between population, natural resources, industrialization, pollution, and quality of life. Three of Forrester’s students further refined World2 on behalf of the Club of Rome. The result – World3 – was published in the 1972 edition of the book The Limits to Growth. The authors develop a number of scenarios based on the model computing state variables every 6 months between 1900 and 2100. Three key plots describe each scenario: state of the world, material standard of living and human welfare and footprint. World3 is available in a number of modern SD packages such as Vensim or Stella and is still used as a basis to work out global dynamics. Dolores Garcia, an independent researcher based in Brighton, UK, started from World3 to produce a revised version of the model incorporating feedback loops related to climate change and energy. While World3 handles only non-renewable resources, Gracia refined for instance energy supply and calculated the progressive decline of EROEI (Energy Returned On Energy Invested) over time for non-renewable energies.

Other integrated global models are listed below.

Integrated Model to Assess the Greenhouse Effect was developed to provide a comprehensive overview of global climate change; it implements the OECD’s Driver-Pressure-State-Impact-Response (DPSIR) approach into a global model by integrating a number of different models for each sector; the latest version has been used to facilitate science-policy dialogues
International Futures simulator was initially developed for educational purposes then used as a policy tool; it was used to assess the attainment of the Millennium Development Goals outlined by the UNDP in 2004; the sectors of the model include population, economy, agriculture, enery, social and international politics, education, environment and technology
Dynamic Integrated Climate and Economy is one of the simplest integrated global models which incorporates a basic treatment of climate change; in DICE, the economy is fairly decoupled from the environment
Tool to Assess Regional and Global Environmental and heath Targets for Sustainability consists of the following sectors: population and health, energy, land, food and water; each sector is based on a DPSIR model and integrated via a socioeconomic scenario generator in which policy responses are reflected
Global Unified Metamodel for the BiOsphere was meant to simulate the earth system and assess the dynamics and values of ecoysystem services; it is the first model to integrate the dynamic feedback between technology, economy, and ecosystem goods and services
Feedback-Rich Energy-Economy explores feedback processes in a climate change context and can be parametrized to reproduce DICE; this model has a rich feedback structure while allowing optimization and uncertainty analyses with reasonable computing resources
Climate Rapid Overview and Decision Support is based on the work behind FREE and is a simulation that helps users understand the long term climate impacts of scenarios to reduce greenhouse gas emissions

System dynamics also found its way into games such as the popular SimCity. Maxis released the first version of the game for PC and Macintosh in 1989. In SimCity, you are the mayor of a city which you have to grow by making supply decisions across a number of sectors. The model matches your decisions against the demand which is calculated by the computer at each round. The model is a black-box and assumes that land is infinite. Taxes are incorporated with its own delays to further complicate decision-making.

System 4 And Shared Models

Stafford Beer, the man behind the Cybersyn Project, defined a viable organization as one able to adapt and maintain an independent existence as it co-evolves with its changing environment. Beer devised a model which identifies key structural requirements to sustain this independent existence. One of the requirements requires the organization to be able to sense the outside and then i.e. perceive threats and opportunities in its environment and devise measures to avoid or take advantage of these changes. In addition to environmental sensing, System 4, as it is called by Beer, also manages a model of its organization and its relationship with the outside world. That was the purpose of the SD model CHECO in Project Cybersyn. CHECO had become the shared model used by Allende’s team to manage the organization but how did they devise this model? Who was involved in the modeling? Was there a process to change the model over time? A shared model would most likely be handled as a living artifact which allows members of the organization to refine its ability to model perturbances and their effects.

Participatory System Dynamics

SD modeling requires a transdisciplinary approach to learning. Transdisciplinarity implies crossing boundaries and inter-sectoral approaches that provide individuals with tools to sense and adapt to changes taking place around them. While SD models produce such powerful tools, they remain complex artifacts usually mastered by few analysts. Below is a model walkthrough (approx. 20min) by Tom Fiddaman, author of FREE and senior modeler at Ventana Systems. As an educator in modeling, he has been experimenting with screencasting formats and posted an SD model of Nathan Forrester which represents a synthesis of Macroeconomics.

This model represents only the economic sector and you imagine the challenge when adding more sectors as done in all models shown above. Simplification is not the answer either. As Ross Ashby said, only variety absorbs variety and the complexity of a world system cannot addressed by cutting corners… How to overcome this communication hurdle and achieve that a model becomes shared among stakeholders?

Krystyna Stave is associate professor of environmental studies at the University of Nevada, Las Vegas since 2003. She has gathered some experience in participatory system dynamics modeling. In a paper published in Sustainability 2010, she makes the case for an analytic-delibration approach in sustainable development management. She promotes the facilitation of participant learning, group model building and mediated modeling. This approach seems very interesting with regards to obtaining a concensus on a shared model and provides the basis for keeping the atifact alive through future adaptations. In her report, Stave presents four cases – water supply and demand, municipal solid waste management, mobility and air quality, land use / transportation / air quality – which vary in participation from none to high. All models leverage the capabilities of modern SD modeling software by presenting the users an easy-to-use dashboard with the problem at hand, the variables that can be tweaked and the results obtained. How could an approach be further expanded and how could people be involved to further refine their shared model of the environment?

Crowdsolving: The Next Frontier?

Jeff Howe of WIRED magazine published Crowdsourcing in 2008. In the book he goes beyond WikiPedia and talks already about NetFlix Prize, Dell IdeaStorm and InnoCentive challenges. All these initiatives highlight the potential of cracking product management and complex problems by exposing them to many eyeballs. This approach is already common in citizen science which I follow already with great enthousiasm as it teaches us how to package scientific protocols for use by the masses. Crowdsolving problems seems the new frontier where a time-factor comes into play and organizations seek the power of the crowd to solve important problems in a given timeframe. This was shown recently with the DARPA Network Challenge. The challenge required teams to provide the coordinates of ten red weather balloons placed in different locations across the US. The winning strategy involved a recursive incentive mechanism which allowed the MIT Team to complete the challenge in approximately 8h52 excluding a prep phase involving the recruiting of 4.400 volunteers in 36 hours.

Interesting research is currently under way by people sich as Luis von Ahn or Panos Ipeirotis to understand the logic behind crowdsourcing and incentive-compatible mechanisms required to meet multiple and conflicting goals. See for example the deck of Panos Ipeirotis on Managing Crowdsourced Human Computation presented at WWW2011. These approaches can be applied to SD modeling as well. What if a Kaggle competition was set up to refine an SD model in pure transdisciplinary style that would make Robert Axelrod proud? Axelrod had crowdsolved a first winning strategy for the iterative Prisonner’s Dilemma tournament back in 1979.

Sustainability to me, has an essential relation to the concept of resilience. Resilence in turn requires the ability to resist perturbances and return to a stable equilibrium. As the viable system model indicates, there are clearly identified requirements to ensure resilience. One of these is the ability to work out future problem areas based on a shared mental model of our environment. This model is evidently complex with unknown structures, time delays, impossibilities, and selected information availability. I believe the SD approach continues to be appealing and can benefit from advances in simulation software as well as modern participatory mechanisms such as those based on crowdsolving.

Creative Commons License This material is licensed under CC BY 3.0

Back to posts