Thinking in Systems: Thermostats

Recommended conceptual framework

As I’ve probably mentioned before, Donella (Dana) Meadows’ book Thinking in Systems is a highly recommended personal favorite.

And — she uses thermostats as an excellent illustration of how complex systems can be simplified into an easier-to-grasp framework.

Figure from “Thinking in Systems” by Donella Meadows

The left side of the figure above shows a “stock-maintaining, balancing feedback loop.” If you were to just look at this section, the inflow of heat, you would see room temperature that changes to match the thermostat setting. With this balance finally met, the furnace would turn off.

“However, this is not the only loop in the system. Heat also leaks to the outside,” writes Meadows. The right side of the figure, the outflow of heat, also shows a balancing feedback loop. Unintended flows cannot be ignored.

So what happens when these two loops operate at the same time? It depends on which feedback loop dominates the system. If heating capability and insulation are sufficient, then the heating loop will dominate. If more heat is leaked than the furnace can contribute, then the cold loop will dominate.

All of this seems fairly straightforward. However, what *should* seem straightforward (but is ignored too often) is the behavior of competing feedback loops. Meadows offers an example. “It’s like trying to keep a bucket full when there’s a hole in the bottom.” She continues, “We are trying to keep the room warmer than the outside and the warmer the room is, the faster it loses heat to the outside. It takes time for the furnace to correct for the increased heat loss.”

A simple, simple, observation. That we forget all the time.

Think about your pantry. You want to maintain a healthy stock of M&Ms for guests who visit your place, hankering for a hospitable snack. HOWEVER, the colorful bursts of chocolate have a seductive control over you, the host. You eat them anytime you walk into the kitchen, depleting your stock. And, Amazon Prime isn’t available in your neighborhood (yet), so “you can’t instantly order new stock to make up an immediately apparent shortfall.” What’s a host to do?

Meadows offers some general principals.

Point #1. “The information delivered by the 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,” writes Meadows.

“A person in the system who makes a decision based on the feedback can’t change the behavior of the system that drove the current feedback, the decisions he or she makes will affect only future behavior.” — Donella Meadows

This seems pretty obvious once you see it, typed out in big font. But in daily practice, this is a truth we easily ignore. My overconsumption of M&Ms today doesn’t impact my consumption of M&Ms yesterday… it only impacts my future intake. And yet, too often I linger on my past deficiencies rather than taking steps forward.

Ultimately, this means that there will be response delays. “A flow can’t react instantly to a flow. It can react only to a change in a stock, and only after a slight delay to register the incoming information.” So, a flow must wait until resulting data aggregates in the form of a stock — the memory of change in flow— before it can respond.

We see this behavior manifest in many other situations, from assessing credit card borrowing to hiring enough high caliber employees to account for folks who quit during your hiring process. It’s the tricky fill the bucket while the bucket leaks scenario.

It’s the percolation of time.

Point #2. “Suppose you knew nothing at all about the thermostat, but you had a lot of data about past heat flows into and out of the room. […] You could find an equation telling you how those flows have varied together in the past.” This equation would hold… as long as the inflows and outflows remained constant.

But, change the structure, change its behavior. “Your equation would hold only until something changes in the system’s structure — someone opens a window or improves the insulation, or tunes the furnace, or forgets to order oil.” Or manually closes a damper.

Which leads us to the most complicated part about supposedly simple thermostat systems. Response delays, while not super easy, seem manageable... leaky buckets and all. There are predictive behaviors of people and systems that we can account for. It’s the introduction of unpredictable inflows and outflows to the system that gives rise to messiness. Messiness to manage.

So, what are ways to counter this unpredictability?

A Stock with Two Competing Balancing Loops — A Thermostat

Author: Donella Meadows



A thermostat is the classic example of a balancing feedback loop. Its purpose is to keep the system stock called “temperature of the room” fairly constant near a desired level. Any balancing feedback loop needs a goal (thermostat setting), a monitoring and signaling device to detect deviation from the goal (the thermostat) and a response mechanism (the furnace and/or air condition, fans, pumps, pipes, fuel, etc).