Key Ideas

• Everything we know about the world is a model 🌱. Our equations, our computer programs, our statistical analyses, they are all limited models to a complex reality. It is when we forget this, that we get into trouble.
• Deduce the purpose of a system by observing its behavior 🌱. Purpose is deduced from behavior and actions, not from rhetoric and stated goals. We won’t be fooled by what a system is meant to do if we look at what its behaviors are.
• A stock is the memory, or current result, of the changing flows of a system.
• Ask yourself: If A causes B, can it be that B also causes A?
• The most common struggle is not to find the leverage point but which way to pull.
• Put your models and relationships in paper, then evaluate how accurate they are. Your thinking will become clearer both by externalizing your thoughts and by iterating on them.

Introduction: The System Lens

Manipulating a system may suppress or release a behavior that is latent within the structure. The change may not be caused by the manipulation itself.

Picture pushing a slinky down the stairs. The push started the motion but its structure keeps it going. The behavior of a system cannot be known just by identifying its elements.

Similarly, we can’t understand a system by looking at its parts individually.

Chapter One: The Basics

A system must have three things: elements, interconnections, and a purpose.

How to know if you are looking at a system: - Can you identify the parts? - Do the parts affect each other? - Do the parts together produce an effect that would be different from the effect of each part on its own? - Does the behavior persist over time?

Deduce the purpose of a system by observing its behavior 🌱. Purpose is deduced from behavior and actions, not from rhetoric and stated goals. We won’t be fooled by what a system is meant to do if we look at what its behaviors are.

A group of agents with the same purpose can give rise to a meta-purpose that is not intended by any of the parts, and may even be counterproductive to the agents’ purpose themselves. That is how the tragedy of the commons happens.

Levels of Change

• Changing the elements of the system has little to no effect.
• Changing the interconnections of the system has a crucial effect.
• Changing the purpose of the system has a critical effect.

Stocks and Flows

A stock is the foundation of any system. It is the memory, or current result, of the changing flows of a system.

A flow represents the change in a stock over time.

The human mind tends to focus on the stocks of a system more than its flows. Furthermore, it tends to focus on the inflows more than on the outflows.

Stocks allow inflows and outflows to remain independent of each other, so they can be out of balance for some time.

Feedback Loops

If you see a mechanism that is behaving the same over time, there likely is another mechanism making sure it stays like this. That is a feedback loop.

There are two types of feedback loops: balancing and reinforcing.

Balancing feedback loops maintain order while reinforcing feedback loops create chaos 🌱. A balancing feedback loop has a goal and it tries to move the current state closer to it. It seeks stability and it is resistant to change.

A reinforcing feedback loop generates more stock the more stock there is and it creates exponential growth. It is found when a stock can reproduce itself, directly or indirectly.

Reinforcing feedback loops happen when the winner is rewarded with something that enables them to win more 🌱. If the resource required to win a competition is finite, which it usually is, and the winner is awarded this resource, the result is the elimination of all but a few competitors. Therefore, the reinforcing feedback loop is a selective mechanism.

Feedback loops cause non-linear behavior in a system. The output of a system is a combination of the input and the output itself.

This raises the question: “if A causes B, can it be that B also causes A?”: - If poverty causes low literacy rates, can it be that low literacy rates also cause poverty? - If exercising leads to a healthy diet, can it be that a healthy diet leads to exercising? - If trust sparks vulnerability, can it be that being vulnerable also sparks trust?

Chapter Two: The Systems Zoo

The … goal of all theory is to make the … basic elements as simple and as few as possible without having to surrender the adequate representation of … experience. Albert Einstein.

Feedback loops cause changes in a future state based on a past state 🌱.

It is crucial to remember that there are always delays in responding to changes. Feedback loops will react only after certain time and with respect to a past state. If the incoming changes are slight outdated, the resulting reaction will be too. That is one of the flaws in economic models that assume that a change in price will instantaneously be faced with a change in demand. The market takes time to react to a change in price.

Aim for higher goal so the balancing loop lands in the intended place.

Example: A thermostat trying to keep a room warm will always have a delay in the activation of the heating system. For this reason, thermostats are set to temperatures slightly above of the desired one.

A balancing feedback can produce oscillations if its reaction time is too short 🌱. Since any system has delays in its feedback loops, if it is reacting too quickly, it is not allowing the system to stabilize.

A Population Model

The combination of a reinforcing feedback loop with a balancing feedback loop on a single stock represents how a population or an economy works.

The main thing to look out for is a dominance shift, that is, when one of the feedback loops overtakes the other one. This means there are three states: growth, stability, and decline.

Dynamic systems studies are focused on the simulating the what if’s and not with predicting the future.

The central question of economic development is how to keep the reinforcing loop of capital accumulation growing faster than population growth. Otherwise, there will be less to share.

Two-Stock Systems

In a resource-dependent system, the capital growth rate is a high-leverage parameter 🌱.

Any physical system with a reinforcing loop is going to run into some kind of stock constraint which will then act as a balancing loop either by strengthening the outflow or weakening the inflow.

Non-renewable Inflow Stock

A non-renewable stock is exploited by a capital-like stock.

Example: oil extraction

The depletion of a non-renewable inflow stock causes costs to rise as the resource is extracted, making the last parts more expensive than their marginal revenue.

The stronger the reinforcing loop of capital growth (the capacity to extract the resource) the faster the resource will deplete and the harder capital will fall.

Another factor is the size of the resource. The larger the resource, the longer capital can grow and, since its growth is exponential, the harder it will fall.

If we keep capital growth small, then the rate of extraction allows for a stable depletion of the resource with respect to the capital.

Renewable Inflow Stock

A renewable stock is exploited by a capital-like stock.

Example: fishing

The reinforcing loop of the exploited stock usually has an upper bound and a lower bound.

If the number of fish in the area is too high, the ecosystem won’t be able to support more fish.

If the number of fish in the area is to low, they won’t be able to restore their population.

Note that fishing up to certain point increases reproduction because resources are freed up.

There are three non-linear interconnections of the resource: the price, the regeneration, and yield per unit.

If a system is to avoid going over the critical point of a renewable resource, the balancing feedback loop has to constrain capital growth fast enough so it doesn’t overexploit the resource.

Non-renewable resources are stock-limited. The more the resource, the longer extraction can go on but the faster it depletes.

Renewable resources are flow-limited. They can support unlimited extraction but it has to be linked to regeneration rate.

Chapter Three: Why Systems Work So Well

Resilience

It is the capacity of a system to bounce back to normality after a perturbation.

Seek to be resilient to changes and not stable over time 🌱. People often sacrifice resilience for stability since the benefits of resilience are often not visible and people fall back to feeling safe.

Resilience is about having stabilizing forces that keep you at the desired level and that those forces still work when you deviate significantly.

Self-Organization

Systems that can change their rules and parameters have an advantage over more rigid systems.

Similar to resilience, we tend to sacrifice self-organization for short-term productivity by specializing.

With relatively simple rules, self-organization can happen.

Example: fractals. Human lungs maximize surface area (the size of a tennis court) because they are built following only fractal rules.

Example: evolution. A species can adapt to its environment because its traits self-organize to maximize survival following only mutation and selection rules.

Hierarchy

Complex systems can evolve from simple systems only if there are stable intermediate forms 🌱.

Hierarchies reduce the amount of complexity that any subsystem has to deal with. Similar to encapsulation in Software Engineering.

Hierarchies appear from the bottom-up. The agents involved figure out that there may be a better way to self-organize and create a layer above them. Top-down solutions rarely have perspective on the ground-level details.

Example: hierarchies appeared from the cell, to the organs, to the body, to the species, to the ecosystem. The ecosystem did not design all the way down to the cell.

If the main system has little control, a subsystem’s goals can dominate the hierarchy. It is called suboptimization and is also known as Tragedy of the Commons.

Too much central control is also detrimental to a system. Subsystems should have the ability to self-organize and function without tight control of the main system. Self-Determination Theory comes to a similar conclusion, placing autonomy as one of its three components.

Chapter Four: Why Systems Surprise Us

Everything we know about the world is a model 🌱. Our equations, our computer programs, our statistical analyses, they are all limited models to a complex reality. It is when we forget this, that we get into trouble.

Levels of Analysis

There is a quote that goes: “Small minds discuss people. Average minds discuss events. Great minds discuss ideas.” Something similar happens in systems analysis.

• Event-level analysis is noisy and superficial. It only tells us the results of a system.
• Behavior-level analysis gives us a better picture of what is going on and we can dare to do near-term predictions but long-term forecasts are going to bite us.
• Structure-level analysis is where we go deep into what are the forces powering a system and we have a bigger picture of it.

If we are stuck looking and fixing the results of a system, we won’t know how its inherent structure is causing them.

Non-linearities are the chief cause of dominance shift. For example, when a pest is growing faster than how predators are growing, they overtake them.

Changing time horizons causes previously inconsequential boundaries to become influential. For example, if we are looking at car manufacturing and looking over hundreds of years, metal mines start to influence the system.

The horizons in a system are always made up, they are there for simplifying and not for being accurate. Where a system begins and ends is completely up to us.

We are constantly acting through bounded rationality because we don’t have complete information. This may make some decisions seem sub-optimal while they were rational given their circumstances.

Chapter Five: System Traps

Policy Resistance

There are systems with multiple players trying to pull levers to their own interests at the same time.

Example: The drug world.

Stricter policies cut down drug supply, which raise prices, which gives resources to drug lords to figure out how to increase supply.

Solutions

• Seek harmonization of goals.
• Letting go to focus on amending the larger goals of the system.

Tragedy of the Commons

It happens in systems where users need to increase the usage of a shared resource but the consequences of doing so are not felt by them. There is a missing or heavily delayed balancing feedback loop.

Example: Industrial waste.

Without regulation, companies could dispose their industrial waste in a way that affects resources used by others, like air and water. They are not affected directly, or in proportion, by their actions, so they are not compelled to change their practices. They don’t directly breathe the air they contaminate.

Solutions

• Educating and exhorting the users through moral systems and social pressure.
• Privatizing the commons to re-establish the feedback loop.
• Regulate the common with a central authority who polices the resource.

Drift to Low Performance

A declining performance of a system makes the users adjust down the goals due to pessimism.

Actors tend to have a negativity bias on bad results and disregard positive ones.

The main problem is that adjustments to the system come to from the pessimistic perception so they end up being half-efforts.

Example: Declining company morale.

Failed projects and declining revenue cause companies to aim for easier goals. Next year, after not meeting those goals, they adjust down the next ones, and so on. The company keeps drifting into mediocrity.

Solutions

• Keep goals absolute 🌱.
• Make goals sensitive to best performances, so you are always improving.

Escalation

A reinforcement feedback loop appears because the state of a system depends on improving over the state of another system.

Example: An arms race.

The military power of a country depends on the perceived military power of another country. Both keep investing on their military to match and overtake the other.

Solutions

• Not entering the game.
• Cease to participate.
• Negotiating.

Success to the Successful

The system rewards the winner with a resource that gives them the ability to win again.

Two species cannot live in the same ecosystem and depend on the same resources. That means that every species in an ecosystem has diversified its dependencies well enough to be unique.

Example: (Unregulated) Capitalism.

The more capital you have, the more you can produce, the more you can earn. You invest those earnings in more capital, engaging the feedback with greater effects.

Solutions

• Diversification so losers of the game can start another one.
• Strict limitation on the fraction of the pie a winner can hold.
• Policies that level the playing field.
• Policies that reward success in a way that does not go to the next round.

Addiction

Due to a gap between the perceived state of a system and its desired state, an intervenor implements a measure to bring it to its desired state. This intervention undermines the system’s capability of restoring itself, giving rise to addiction 🌱.

Withdrawal is the confrontation with the real state of the system.

Example: Modern medicine.

Instead of making the individual responsible for their health, modern medicine covers the deficiencies caused by a poor lifestyle. The individual depends on modern medicine to fix problems caused by their own decisions.

Solutions

• Shift away from symptom-relieving policies.
• Implement measures that strengthen a system’s ability to restore itself.
  • Why are the natural correction mechanisms failing?
  • How can obstacles to their success be removed?
  • How can mechanisms to their success be more effective?

Rule Beating

The conterproductive action that is done to circle around a rule but keeping the appearance of it being followed.

It violates the spirit of the law while preserving the word of the law.

Example: Budget spending.

Institutions or departments come up with pointless spending at the end of the year to finish their budgets, so they don’t lose them next year.

Solutions

• Redesign the laws in order to release creativity in the direction of following the rule.

Seeking the Wrong Goal

The success measure of a system does not reflect its values.

It is almost the opposite of rule beating because following the rule is in itself doing the damage.

Example: GDP.

GDP does not reflect the well-being of a country’s citizens. It is merely the movement and production of goods, however inefficient and wasteful.

Honorable mentions: Money spent in military as national security, Standardized tests as education level.

Solutions

• Reevaluate goals not to aim only for effort but also for result.

Chapter Six: Leverage Points

The most common struggle is not to find the leverage point but which way to pull.

12. Numbers and Parameters

It doesn’t matter how you much you open or close a faucet if the plumbing remains the same.

There is a big grey area where the parameters of a system produce the desired results.

11. Buffers

Bigger buffers with smaller flows produce more stable systems.

There are often physical constraints to them so they have low leverage.

10. Stock-and-Flow Structures

This involves changing the underlying structure of a system like the plumbing of a house, the information channels of a government, the speed limits of a city.

Careful design of a system is crucial because, once it’s finished, it’s hard to change.

9. Delays

How long a system takes to react can be hardly changed since that depends on its structure and things take the time they need, otherwise they can be a high leverage point.

8. Balancing Feedback Loops

This means intervening to increase the effectiveness of a system’s self-preserving mechanism. Examples:

• Well-being: nutrition, exercise, socializing.
• Fair markets: protection to whistle-blowers.
• Environment: Pollution taxes.

7. Reinforcing Feedback Loops

Systems with unchecked reinforcing feedback loops will destroy themselves, that’s why there are so few of them. Keep an eye open for the selection bias of these systems.

It is better to slow down growth than fixing the consequences of growth. In other words, it is better to slow down when driving too fast than building a better brake system.

6. Information Flows

This means creating a new feedback loop, which makes consequences of actions directly affect the actors. Make agents accountable for the feedback loops they create.

This is a high leverage point for systems suffering Tragedy of the Commons.

5. Rules

Rules give rise to incentives and punishments, which agents then act on.

Mikhail Gorbachev changed two levers: the information flow and the rules of the system, and the Soviet Union saw enormous change.

Having power over the rules means having power over others’ behavior.

4. Self-Organization

If a system itself can change the rules, then it will be resilient to internal and external changes.

A system that is restricted from evolving and changing is destined to crumble when circumstances change.

3. Goals

If the goals of a system contradict any of the rules, information flows, or even the self-organization nature, it will override these to fit them.

Goals have to remain relevant at the broader level, never putting the unit before the whole. Just like one company’s future is not the priority over a country’s economy.

2. Paradigms

These are the ideas the give birth to the systems. Human greed can be channeled for the greater good. Capitalism tries to do that.

Paradigms represent the world in a particular way. They are beliefs and assumptions of how things should work.

This is the leverage point of ideologies. Since they ingrained in the actor’s minds, they are hard to override. But once paradigms shift, huge changes happen.

To defeat old paradigms, keep pointing at the deficiencies and failures, and introduce people who bring new paradigms.

1. Transcending Paradigms

Realize that paradigms are a limited view of the world and do not encompass the whole truth. There is no paradigm that is completely correct. They are all approximations of reality.

Keeping ourselves unattached from paradigms lets us easily adopt new ones as circumstances change.

Chapter Seven: Living in a World of Systems

This chapter sums up the wisdom in this book. If you are looking at getting all the wisdom from Donella without diving into too many details, read this chapter and chase the concepts that sound interesting.

Outputs of a systems cannot be controlled. Systems can be designed to make the desired behavior likely to happen.

Similarly, the future cannot be predicted but systems can help us shape it and provide the ground for the future we would like.

Observe Before Intervening

Focus on the facts of a system, instead of starting with theories. See what works, what are the feedback loops, who are the actors, etc.

Look at the history, before building a solution.

Test Your Models

Put your models and relationships in paper, then evaluate how accurate they are. Your thinking will become clearer both by externalizing your thoughts and by iterating on them.

Emphasize Information

Most of what goes wrong in systems has to do with incorrect, late, and corrupt information. The outputs of a system should be transparent to its actors.

Use Precise Language

Expanding on the previous point, information should be accurate and descriptive.

We don’t talk about what we see; we see only what we can talk about 🌱.

For example, goals should be defined with precision so the system can see the world through their lens.

Enlarge your language so you can cover the complexity of the systems.

Focus on Quality

Don’t sacrifice important parts of the system only because they are hard to measure. This includes, resilience, stability, and adaptability.

Be a quality selector, follow your intuition of quality, and rely less on noisy signals.

Make Feedback Policies

Feed sewage back to the polluter’s inline. Create immediate consequences to actor’s actions.

Enforce policies so that actors have skin in the game.

Aim for the Whole Good

Remember when hierarchies and systems are there for the bottom layers and not the top. Do not fall for the GDP trap by optimizing for the wrong measure.

Keep the Pareto Principle in mind when looking to improve a system.

Celebrate Complexity

Observe how nature has built itself over time and the patterns it uses.

Create systems that embrace complexity, even if they sacrifice perfection.

Examples: Neural Networks, Persian carpets, coffee.

Preserve Absolute Goodness

Explaining away our mistakes through biology and psychology is an easy escape. Face your mistakes and seek to overcome the faults of your own nature.

Conclusion

Thinking in Systems was one of my favorite books I read on 2020. The writing is easy to understand and it lands abstract concepts into real life examples. It is a great introduction to this world and from there we can branch out into more specific areas of systems analysis.

Thank you for reading.