Thinking in Systems, a Short Review
March 7, 2019
We know models to be untrue, but if they are close enough to reality we can use models to learn, imagine potential futures, and make better decisions. Donella Meadows’ Thinking in Systems is a study of models, and the ways in which seemingly different problems in various settings can be understood using the toolset of stock and flow analysis. In this context, flows represent a rate of change in some unit, while stocks are the accumulation of that unit; the system’s memory of the history of changing flows. Her book argues that common structures arise from the interaction of simple sub-systems, and that by pattern-matching against a few archetypes we can grasp the long term behavior of many systems. She suggests that we should treat complex system structure as the source of behavior, and think of everything else as epiphenomena. Paraphrasing Meadows, we can use the fact that similar feedback loops lead to similar dynamics to simplify the task of understanding the world around us.
Starting with the simplest examples (think a tub being filled with water) Meadows takes the reader through what she calls the “Systems Zoo,” a collection of systems pulled out of context into neat self-contained examples. The conversation quickly moves from tubs and thermostats on to small inventory systems, and eventually to population dynamics and even whole economies. The idea of being able to apply the same abstraction to understand such diverse systems permeates the book, which while published in 2008 was originally drafted in 1993. Reading it made me wonder how engineering education has evolved in the last ~30 years. The generalization I just mentioned was not new to me today, but I remember the power of that abstraction being surprising during my first years of college, where it was an essential part of the engineering curriculum. From how the book is written, it seems like its centrality wasn’t widespread back then, although I have no data to back that claim.
The most useful and applicable part of the book for me was the discussion about boundaries and tradeoffs:
There is no single, legitimate boundary to draw around a system. We have to invent boundaries for clarity and sanity; and boundaries can produce problems when we forget that we artificially created them. When you draw boundaries too narrowly, the system surprises you.
Whatever assumptions we make along the way end up embedded in our models. We draw up artificial boxes and assume everything else away, compressing the world and baking in our biases. This is true in the world of programming, where we constantly make assumptions about time, names, and more, but its also true in economics with homo economicus and perfect information. For example, she brings up the variables most commonly solved for in economic models:
Economics evolved in a time when labor and capital were the most common limiting factors to production. Therefore, most economic production functions keep track only of these two factors (and sometimes technology). As the economy grows realteive to the ecosystem, however, and the limiting factors shift […] the traditional focus on only capital and labor becomes increasingly unhelpful.
The decision to use labor and capital as independent variables is so ingrained in the discipline that we don’t even question the exclusion of other metrics. Meadows invites us to reconsider many of these given ideas, and insists that good models should acknowledge the trade-offs inherent in these forgotten decisions. Other fields are also plagued with similar problems of implicit and forgotten assumptions. The book mentions a few in passing, then takes a turn and deep dives into sustainability and ideas around the limiting factors of growth.
While I enjoyed reading it, and would recommend it to people who haven’t studied systems thinking before, the book left me somewhat disappointed, wanting more. Its tagline as a “primer” is apt. The little math there is appears in the appendix — and even then it is simplified — and the high level explanations are sprinkled with simple diagrams. If you have recommendations for other books in this area that go deeper into the technicalities, please let me know.
Decision making in complex systems is hard, especially when aiming for long term goals under uncertain conditions. Meadows’ prescription of thinking about systems hollistically is the only way to tame unpredictability. Systems thinking is about finding leverage points and determining bottlenecks. If through analysis we can establish which aspects of a system are most likely to change the system’s behavior, we can develop solutions around those critical points, take advantage of the system’s built-in feedback loops, and move a step or two closer to our objective.
Photo: by me, previously posted on SF Cityscapes
Want to see more articles like this? Sign up below: