Summary
I have yet to see any problem, however complicated, which, when you looked at it in the right way, did not become still more complicated. -Poul Anderson
Donella Meadows was an integral part of a group out of MIT who pioneered the art of “systems analysis”. How does one analyze a system? It starts by defining what the essence of a system is. According to Meadows a system is a group of things material or immaterial that interrelate to cause a behavior which persists through time. In plain English, behaviors are things we observe like floods or forest fires. Systems are the relevant variables that have an effect on the observed behavior, like rain or ground cover. In analysis, behavior is typically the easy thing to observe while defining the variables that constitute the system underneath can be quite a task and isn’t solved but often requires flexibility and good judgement.
Getting to the core of the book we have two central concepts for systems analysis; stocks and flows. A stock is a quantity of something that can be increased and decreased, those changes are called flows. The canonical example in the book is that of a bathtub, you have a stock which is an amount of water in the tub. There is a flow from the faucet which increases the stock of water in the bath tub while the drain represents an additional flow that decreases the level of water in the tub. Even with this simple system, one stock and two flows, you can create some interesting relationships like matching the flow rate of the faucet to that of the drain to maintain a certain water level. This also highlights the earlier point about defining systems being somewhat creative, for example, we could have also included evaporation as an additional negative flow on the water level and on and on.
Another tool this book offers is the ability to analyze the types of flows that can operate on a stock. There are two main types of flows:
Balanced Loops: This is a flow that attempts to pin a stock at a desired level. A house thermostat is a good example of this, you set a desired level and the thermostat runs your furnace and air conditioner in an attempt to stay at that level. The larger the gap between actual and desired is, the harder the system will “push” towards the desired level.
Self-Reinforcing Loops: These are flows that grow exponentially, cell division is a great example, the stock is cell count, with every cell division there exists more cells to divide and so on.
In the real world, systems don’t have boundaries, everything is connected to everything else, yet boundaries are necessary for analysis and so choosing boundaries wisely becomes the most difficult and important part of analysis. When these boundaries are drawn interactions between systems can be observed. Meadows outlines the following system interaction archetypes:
Policy Resistance: This is when different systems have opposing goals and react to each other with negative patterns. This tends to cause interventions to a system to be canceled out because any adjustment to one system causes the other system to react in an attempt to cancel out the change. A great example here is England’s window tax (1696), the result was that many homeowners either built houses with fewer windows or boarded up existing windows to pay less taxes.
The Tragedy of the Commons: This is when there is a shared pool of resources but the penalty of misusing the resource is also shared. This creates a perverse incentive for the individual to misuse the resource before anyone else can because the marginal value is higher to the individual than the cost which is borne by all. There are many of these examples, polluting, overfishing, poaching are a few that come to mind.
Escalation: This is when the desired target of a stock is set in an attempt to match a different stock’s value. This is similar to Policy Resistance but instead creates runaway growth in a stock instead of deadlock. Nuclear weapons make for a good example where every nation feels the need to match other countries’ nuclear deterrence through treaties or production.
Success to the Successful: This is when any increase in a stock results in an increase in rate of growth of that stock. Compounding interest is a common example of this type of interaction.
Addiction: This is when a system’s stock value becomes dependent on another system’s output. Oil is a good example here, the discovery of oil greatly increased the productive capacity of humanity at the same time as making us addicted to energy which, for now, is largely provided by oil.
Meadows also prescribes some ways of trying to work with each of these types of related systems interactions while also noting that reckless interventions can result in unpredictable distortions to behavior.
At this point in the book Meadows switches gears from description to prescription.
Meadows argues that identifying the system causing the behavior makes attempting to tune the parameters of the system tempting, but that this often is not the highest point of leverage, the problem is by definition systemic and requires a new system. That being said, sometimes limiting factors can be found in a system that are acting like a dam, once removed the system can rapidly approach a more healthy equilibrium.
Another key principle of Meadows’ approach is a rejection of infinite growth. If an entity is physical then it is surrounded by limits. For example a virus can spread quickly but eventually its growth is limited by available hosts and so the spread will eventually slow.
Thoughts
The first half of this book is excellent, it has inspired several software projects I’m working on to see how far I can push some of the assumptions in this book. Meadows’ categories of stocks, flows, feedback loops, and system boundaries are clear and helpful ways of classifying things.
This was one of those books where I was taking a lot of notes because most chapters made very interesting claims that I wanted to remember to dig through later. Often my issue with these types of frameworks is figuring out how much of it is linguistic versus actually adding value. When presented with new terms you are tempted to use them everywhere but soon find out they come with their own baggage that needs to be handled carefully.
There is a very distinct point in the book where Meadows shifts gears, when she does the quality of the book plummets. The thesis that starts this shift goes something like “maybe the highest leverage point is letting go of control and learning to dance with the system”. The charitable read is that instead of working against the systems you should use their momentum to accomplish your goals. Meadows complicates this reading though as she spends the next sections lecturing the “control freaks” of the world who would see this framework only as a way to increase their control over the universe when the universe should be participated in. While I understand the critique, it cuts against the reason for the book’s existence. The only reason to do systems analysis in the first place is to understand a system so one can hope to effect positive change on the system, otherwise what is the point? Of course I also agree that the world is full of unintended consequences and often times even the best intentioned interventions cause more problems than they solve, but the whole point of this book so far has been trying to push back the veil of ignorance to decrease the incidence of these unintended consequences.
To make things worse, many of her prescriptions are based on standard popular talking points that sound good but aren’t borne out by history or experience. A few of the worst offenders:
- Wars have become more irresponsible since the leaders are no longer forced to be in the frontlines.
- Interest is one of the worst ideas ever as it sacrifices long term stability for short term gain.
- Politicians should embrace the complexity of the universe and be more willing to admit the limits of their knowledge.
All of these observations share the same aspect of having popular appeal without the merit to back the popularity up.
Unfortunately, the end colors the beginning and so whenever I think of this book, I will think of its ending. Even so, the ideas presented in the first half were so helpful that they more than make up for some of the more sloppy applications made in the last half of the book.