Systems Thinking

I have yet to see any problem, however complicated, which, when looked at in the right way, did not become still more complicated

-Poul Anderson

Systems Thinking is a hot topic in the world of evaluation. I’ve read a fair number of articles that talk about systems thinking (particularly around applying the concepts to evaluation), but figured it was time that I did some dedicated reading on the topic, as I find that sometimes reading secondary sources means that some components get missed or connections aren’t quite clear. The book that I chose to read is Thinking in Systems: A Primer (2008) by Donella H. Meadows (edited by Diana Wright), which was recommended to me *and* which is available online for free from this link. 🙂

Here are my notes from Part 1 – System Structures and Behaviour:

• “A system is a set of things—people, cells, molecules, or whatever—interconnected in such a way that they produce their own pattern of behavior over time. The system may be buffeted, constricted, triggered, or driven by outside forces. But the system’s response to these forces is characteristic of itself, and that response is seldom simple in the real world.” (p. 2)
• some implications:
  • “Political leaders don’t cause recessions or economic booms.
    Ups and downs are inherent in the structure of the market
    economy.”
  • “Competitors rarely cause a company to lose market share.
    They may be there to scoop up the advantage, but the losing
    company creates its losses at least in part through its own
    business policies.” (p. 2)
• We like to break things down into small pieces to try to understand them; we like to draw direct linear cause-and-effect pathways; we want to solve problems by taking action to control the world around us – this is how we are often taught to do things
• But we grow up in a complex world and learn from experience how these complex systems work
• Reductionism and systems thinking are complimentary – just like you can see some things with the naked eye and other things with a microscope (and both things exist), reductionism and systems thinking are different ways of looking at things
• ” A system is an interconnected set of elements that is coherently organized in a way that achieves something” (p. 11): three things a sysmte has:
  • elements
  • interconnections
  • a function or purpose
• systems are wholes that are more than just the sum of the parts – they have “an active set of mechanisms to maintain that integrity” (p. 12)
• Systems “may exhibit adaptive, dynamic, goal-seeking, self-preserving, and sometimes evolutionary behaviour” (p. 12)
• interconnections can be physical flows or flows of information: “signals that go to decisions points or action points within system” (p. 13)
• functions/purposes even harder to see than interconnections – may not be explicitly stated – sometimes have to watch the system to see how it behaves
  • as well, sometimes systems state a purpose (“this government protects the environment”) but doesn’t behave in a way that works towards that purpose (e.g., puts no money nor effort towards protecting the environment). Basically, actions speak louder than words.
  • most systems have a purpose of “ensur[ing] its own perpetuation” (p. 15)
  • purposes need not be intended by humans and, in fact, the “purpose” of a system might be something that no one in the system actually wants (e.g., society produces crime and drug addiction, which no one wants but which results nonetheless)
  • systems can be nested within other systems = purposes within purposes – and these sub-purposes could conflict with other sub-purposes and/or with the overall purpose. To have a successful system, the system’s purpose and the sub-purposes need to be in harmony
• changing elements in a system doesn’t change the system that much (e.g., switch all the players on a hockey team and you still have a hockey team) compared to changing the interconnections (e.g., change the rules of hockey to different rules and you’d have a different sport)
• changing the function/purpose will result in drastic change to the system (e.g., if you changed the hockey team’s purpose from winning to losing)
• A stock =  “the elements of the system that you can see, feel, count, or measure at any given time” (p. 17) – can be physical (e.g., books in a library, water in a bathtub), but  doesn’t need to be  (e..g., your self-confidence)
• flow = actions that change stocks (e.g., filling, draining, being born, dying, deposit, withdrawal). “A stock, then, is the present memory of the history of changing flows within the system” (p. 18)
• dynamic equilibrium: occurs when the inflow = outflow, so level of stock remains constant (even though the “stuff” of the stock is continuously flowing through it)
• “A stock takes time to change, because flows take time to flow. […] Stocks
  usually change slowly. They can act as delays, lags, buffers, ballast, and sources of momentum in a system. Stocks, especially large ones, respond to change, even sudden
  change, only by gradual filling or emptying.” (p. 23)
• these time lags can be a problem (e.g., we can’t make things happen as fast as we want them to because it takes time to build factories, create and distribute technologies, educate people to work in those factories and use those technologies), but it can also give us “room to maneuver, to experiment, and to revise policies that aren’t working” (p. 23)
  • don’t expect things to happen faster than they can happen
  • where possible, “use the opportunities presented by a system’s momentum to guide it toward a good outcome” (p. 24)
• inflows and outflows can be independent of one another (e.g., we don’t have to produce things at the exact rate that we use them) – we have reservoirs for stocks so that we can deal with this (e.g., banks allows us to store money we’ve made until we are ready to spend it. Inventories allow us to produce things at a different rate than the variable rate of customer demand).
• people make decisions based on stock levels (e.g., inventories too high –> cut prices; money in the bank –> purchase and/or investment decisions about what to do with that money) – this is feedback!
• feedback loop: “when changes in a stock affect the flows into
  or out of that same stock”
• balancing feedback loop: works to maintain the stock within a specific range of values; if stock is too high, it works to lower it; if it’s too low, it works to raise it; a.k.a. “goal-seeking or stability-seeking”
• having a feedback loop does not necessarily mean you’ll reach the target though – there can be any number of reasons why it might fail (e.g., response is too weak, too inefficient, too delayed, don’t have enough resources, don’t have the right information feeding into the feedback loop)
• reinforcing feedback loops (a.k.a., amplifying, self-multiplying, snowballing, vicious cycles, virtuous cycles): “generates more input to a stock the more that is already there (and less input the less that is already there”; can “cause healthy growth or runaway destruction” (p. 30-31); can occur whenever an element has the “ability to reproduce itself or grow as a constant fraction of itself” (p. 31) – e.g., populations and economies; growth is exponential
  • doubling time: “time it takes for an exponentially growing stock to double in size = 70 divided by the growth rate (as a percentage); e.g., at 7% interest, money in the bank will double in 10 years (70/7 = 10)
• usually multiple feedback loops occurring at the same time – several loops pulling it in different directions; a flow may fill one stock and drain another
• dominance: “when one loop dominates another, it has a stronger impact on behaviour” (p. 44) – since systems often have more than one feedback loop, it’s important to think about which one is dominate
• dominance can shift as flows change
• systems dynamics models “explore possible futures and ask “what if” questions”‘; “not meant to predict what will happen” (p. 47)
• questions to test the value of a model:
  1. “are the driving factors likely to unfold in this way?
  2. if they did, would the system react this way?
  3. what is driving the driving factors?”

Some Types of Systems

One Stock Systems

1. one stock with two competing balancing loops (e.g., a thermostat)

• when temperature gets too low, heat inflow kicks in
• when temperature then reaches the set point, heat stops
• but heat can also be lost to the outside, and that feedback loop is trying to make the room temperature = outside temperature (e.g., if it’s colder outside, the heat will flow to outside and bring down temperature of the room).
• as the furnace heats the house, it makes the room temperature hotter, which makes a bigger difference between inside and outside, so heat flowing out increases
• thermostat will make furnace come on, but once it hits the set temperature it goes off, and heat keeps flowing out, so furnace kicks on again (but since it takes time for furnace to kick in, temperature will be a bit below the set point).
• a well-insulated house will slow the leak of heat outside, which gets the room closer to the set temperature
• important general principle: “the information  delivered by a
  feedback loop can only affect future behavior; it can’t deliver the information,
  and so can’t have an impact fast enough to correct behavior that drove
  the current feedback” (p. 39)
  •  “there will always be delays in responding” (p. 39)

2. one stock with one reinforcing loop and one balancing loop (e.g., population and industrial economy)

• e.g.,  —births—> [population]–deaths–>
  • if fertility = mortality, population stable
  • if fertility > mortality, population grows exponentially
  • if fertility < mortality, population dies off
•  e.g. economy
  • —investment—>[capital stock]—depreciation—>
  • the greater the stock of capital (e.g., factories, machines) and the greater the efficiency of production (i.e., output of goods/services per unit of capital), the greater the output of goods and services (and thus capital increases – reinforcing loop)
  • but things wear out and become obsolete – the faster this happens, the shorter the lifetime of the capital
  • if investment > depreciation (as it has been of late), economy grows
• “systems that have similar feedback structures produce similar dynamic behaviors, even if the outward appearance of the systems is completely dissimilar” (p. 51) – e.g., even though the economy and population look very different, they both have a reinforcing loop with a balancing loop

3. a system with delays – e.g., business inventory

•  e.g., there is an inflow of deliveries from the factory and outflow of sales
• car dealer wants to maintain a consistent amount of inventory to both have enough cars on hand for the expected sales, plus some buffer because car sales are unpredictable on a day to day basis
• perception delay: the car dealer monitors sales and when they see to be changing, changes order; takes a few days to determine if the change is normal fluctuation or an actual trend
• response delay: doesn’t make the whole change at once – e.g., orders 1/3 of the increase with each of the next 3 orders (again, just to make sure it’s a real trend)
• delivery delay: how long it takes factory to make & deliver the extra cars; not in the control of the dealer
• without any delays, if you saw a 10% increase in perceived sales, you’d up orders by 10% and shift to a new inventory levels that is 10% higher
• but with delays, you get oscillations (e.g., perception delays –> inventory drop as you are selling more cars then expected but haven’t yet replaced them and increased order to make up new sales; delivery delay means inventory continues to drop, so you order more, but then when they come in, your inventory ends up over, and then the same thing happens over again – oscillations)
• dealer doesn’t have timely information to make decisions and physical delays mean she can’t have an immediate effect for an action she takes
• one might think that since delay is the problem, then shortening her reaction time would mitigate the oscillations, but it actually increases oscillations, because she’s overreacting
• people often have counterproductive idea of what “policy lever” to pull and in what direction to pull it, to get the desired effect in the system
• delays are common in systems and “changing the length of a delay may (or may not, depending on the type of delay and the relative lengths of other delays) make a large change in the behavior of a system” (p. 57)

4. A Renewable Stock Constrained by a Non-renewable Stock – e.g., an Oil Economy
   • previous examples did not include constraints (so we could look at the dynamics in a simple way), but in reality there are always constraints (given a finite environment – the inflows have to come from somewhere and the outflows have to go to somewhere)
   • can have resource constraints and/or pollution constraints
   • constraints can be renewable or non-renewable
   • limits can be temporary or permanent
   • eventually there must be some accommodation (e.g., system adjust to the constraint, the constraint adjusts to the system, or both)
   • when the resource stock is nonrenewable, like oil, it means that as you deplete the resource stock, it gets harder and harder to get the next unit of oil, so more capital is required to extract it
   • at the start, you have high profit, so invest in more capital, so you extract more oil, which depletes the resource stock, which in turn makes it most costly to extract more oil, which decreases your profit – eventually you get to the point where you can no longer make a profit by extracting the remaining oil, so you leave it in the ground
   • you can decrease the rate of growth, which will extend the time to depletion (you can get rich quickly or get less rich, but for longer)
   • if price of the resources goes up or if technology decreases operating costs (e.g., new technology to get oil out of the ground for cheaper) will also delay how long it takes before it’s no longer profitable to extract the remaining oil

5. A Renewable Stock Constrained by a Renewable Stock – e.g. Fishing Economy

• like the previous, except that the input is renewable (e.g., fish can produce more fish; or sunlight, which is constantly replenished).
• the more scarce the resource, the more costly it is to harvest it (e.g., fewer fish require bigger boast to go farther to catch them, more expensive sonar technology to find them)
• regeneration rate of fish is dependent on fish density
  • very high fish density – less reproduction (limited by food and habitat
  • as fish density decreases, reproduction rate increases, until it hits some maximum
  • after that point, when fish density get low enough, reproduction rate falls (as density is so low that fish can’t easily find each other or other fish more into their habitat
• three non-linear equations govern this simplified model of a fishing economy
  • price (more scarce fish = more expensive)
  • regneration rate (less reproduction at very high or very low density)
  • yield per unit of captical (efficiency of fishing technology/practices)

• “this system can produce a number of different sets of behaviours” (p. 67)
  • e.g., fish population increases, fishing increases until the population becomes so low that it’s not economically viable to continue to grow the industry –> fishing levels decrease to the point that we have a sustainable equilibrium
  • e.g. 2, fish population increases, fishing increases and develops technology to increase efficiency such that they overshoot equilibrium point and then oscillation occurs around that point (instead of a steady equilibrium)
  • e.g. 3, fish population increases, fishing increases and develops such good technology to increase efficiency such that they overshoot equilibrium point to the point that they have a near complete wipeout of the population and the industry collapses
  • sometimes there is enough of the population left that, once industry collapses, they can repopulate and the whole cycle can repeat (e.g., forestry)
  • however, whether a renewable resource can bounce back depends on what happens during the time when it is depleted – e.g., a small population of fish might be displaced by other species, or lack sufficient genetic diversity to flourish or be vulnerable to pollution or storms
  • also depends on how fast and effective the balancing feedback loop is to stop capital growth

• “The trick, as with all the behavioral possibilities of complex systems, is to recognize what structures contain which latent behaviors, and what conditions release those behaviors—and, where possible, to arrange the structures and conditions to reduce the probability of destructive behaviors and to encourage the possibility of beneficial ones” (p. 72)