Pattern/Systems Thinking/No. 0379
Feedback Loops
Feedback loops are chains of effects that return to influence their source. In cybernetics, positive or reinforcing feedback amplifies a change, while negative or balancing feedback opposes it. These terms describe the direction of feedback, not whether its effects are helpful or harmful.
- Evidence
- Well established
- Read
- 6 min
- Links
- 12 connections
01You've seen this when…
- in life
The shower feels cold, so you turn up the heat. Nothing happens yet. You turn it up again, then jump away when the hot water finally arrives.
- at work
Your support team is overloaded, so agents rush their replies. Customers write back for clarification. The extra messages leave even less time for the next reply.
- out in the world
A city widens a congested road. The faster trip attracts drivers from other routes and travel times. As traffic builds, some of the initial improvement disappears.
02The idea
A speaker sends sound into a room. A nearby microphone picks it up and sends it back through the speaker. If enough sound returns, each trip around the circuit makes it louder. The squeal is a property of the connection between the microphone and speaker.
A feedback loop exists when a change travels through a chain of effects and returns to influence where it started. There are two basic directions:
- Reinforcing feedback strengthens the starting change. More sound produces more sound. Falling customer numbers can also reinforce themselves: fewer customers mean less revenue, which worsens service and drives customer numbers down further. This is reinforcing feedback, also called positive feedback.
- Balancing feedback opposes the starting change. A thermostat detects a falling temperature and switches on the heat. The warming room eventually switches it off. This is balancing feedback, also called negative feedback.
Positive and negative describe the direction of feedback. Either direction can help or harm. A reinforcing loop can build useful skills or deepen a crisis. A balancing loop can maintain a safe temperature or keep an organization stuck.
03Why it happens
We often explain events as chains that end somewhere: workload causes rushed replies, which cause complaints. But the complaints can return as additional workload. Once that connection closes, yesterday’s effect becomes today’s cause.
Several features determine what happens next:
- The return path must exist. Feedback requires a causal connection back to the starting point. Two things rising together leave that connection unproven.
- The strength of the response matters. A microphone can pick up a little speaker sound without squealing. Whether the sound grows depends on how much survives and gets amplified around the circuit.
- The response takes time. A feedback delay can leave you correcting a problem that has already begun to resolve. Repeated corrections then produce overshooting, as with the shower.
- Several loops usually compete. Growing demand may attract investment, which improves service and attracts more demand. But limited staff, space, or money eventually push back. The dominant loop can change as the system grows.
04A worked example
In 2007, the British bank Northern Rock relied heavily on money-market borrowing to fund its mortgage business. When that funding became difficult to obtain, the bank sought emergency support from the Bank of England. News of the support became public in September. Depositors queued outside branches to withdraw their savings.
What it looks like Many customers independently making the same decision to protect their money. Each withdrawal seems small compared with the size of the bank.
What’s actually going on Withdrawals reduce the cash available to meet further withdrawals. Visible queues also give other depositors a reason to worry and join them. Concern produces withdrawals; withdrawals and queues produce more concern. Individual caution helps create the danger everyone is trying to escape, a mechanism related to a self-fulfilling prophecy.
Northern Rock’s funding problem came first. The withdrawal loop amplified the crisis already underway. That distinction matters: identifying feedback should account for the event that started it and the weaknesses that made it possible.
What would have helped Depositors needed credible assurance that their money would remain accessible, backed by the capacity to honor it. The government’s guarantee of existing deposits helped stop the run by weakening the reason to withdraw ahead of everyone else. The bank’s underlying funding model still needed repair. Interrupting an amplifying loop can buy safety while leaving the underlying problem unresolved.
05How to spot it
06What to do about it
- Draw the return path. Use three to six concrete variables: unresolved tickets, time per reply, repeat contacts. Connect each cause to its effect, including the arrow back. Mark uncertain connections and treat the drawing as a hypothesis to test.
- Test the weakest causal claim. Look for a plausible mechanism and check the timing against evidence from occasions when the connection changed. Correlation alone leaves the direction of influence unresolved.
- Interrupt one connection. With microphone feedback, moving the microphone away from the speaker can work better than repeatedly turning the volume down. With rushed support replies, protected time for complete answers may reduce repeat contacts.
- Allow time for the response. After an adjustment, wait long enough to observe its effect before stacking another on top. When waiting is unsafe, use smaller adjustments and closer monitoring.
- Track what builds up. Backlogs, debt, inventory, and fatigue can continue accumulating while headline numbers look stable. Distinguish these stocks from the flows that add to or drain them.
- Check the new loop you create. A subsidy, guarantee, or extra capacity changes incentives as well as immediate conditions. Look beyond the first improvement to the response it invites.
07When it isn’t a feedback loop
A response forms a feedback loop when its effects return to influence where it started. Sending a reminder is a one-way action unless what happens afterward influences the next reminder. In open-loop control, action proceeds independently of the result. Closed-loop control adjusts action in response to the result.
Nor does every reinforcing loop grow without limit. Resources run out, connections weaken, and opposing loops take over. Limits to growth are part of the explanation, not exceptions to ignore.
Balancing feedback also doesn’t guarantee smooth stability. Strong corrections combined with long delays can make a system oscillate or become unstable. Naming the loop is a start; its strength, timing, and surroundings determine its behavior.
08Roots
In 1868, James Clerk Maxwell examined a device that factories depended on: the governor that regulated a steam engine’s speed. In a centrifugal governor, spinning weights moved outward as the engine sped up, reducing its steam supply. The engine’s motion helped control that same motion. Yet a regulator could also make speed fluctuate. Maxwell used mathematics to investigate when governors would settle and when they would become unstable. The mechanism existed long before the modern vocabulary.
Telephone engineering made another use visible. Amplifiers needed to strengthen weak signals without badly distorting them. Harold S. Black developed a way to feed part of an amplifier’s output back against its input. Sacrificing some amplification could make the result steadier and more faithful. His 1934 paper described stabilized feed-back amplifiers: feedback was becoming an explicit engineering method, building on its earlier presence in mechanical regulators.
Norbert Wiener helped carry the idea across disciplines. During World War II, his work on aiming anti-aircraft fire combined prediction and correction to account for moving targets. In Cybernetics in 1948, he connected control and communication in machines with processes in living organisms. His contribution was a shared framework for the existing idea of feedback. It helped researchers recognize the same circular structure in devices and bodies and, eventually, in organizations and public systems.
09How solid is this?
Feedback mechanisms are mathematically described and experimentally demonstrated in engineering and physiology. In social systems, a proposed loop remains a causal hypothesis until its connections, strength, and timing are supported by evidence.
10Connections
- Often confused with Correlation
- Can lead to Emergence
- Includes Feedback Delay, Balancing Feedback, Reinforcing Feedback, Limits to Growth Archetype, Self-Fulfilling Prophecy
- See alsoOpen-Loop vs. Closed-Loop Control, Stocks and Flows, Gall’s Law, Second-Order Thinking, System Boundary
+ 2 more in the list
11Origin and sources
Feedback mechanisms predate the modern term. James Clerk Maxwell analyzed governors in 1868; Harold S. Black developed negative-feedback amplification; Norbert Wiener unified and popularized feedback through Cybernetics (1948).
- [1]Maxwell, J. C. (1868). On governors. Proceedings of the Royal Society of London, 16, 270–283.
- [2]Black, H. S. (1934). Stabilized feed-back amplifiers. The Bell System Technical Journal, 13(1), 1–18.
- [3]Wiener, N. (1948). Cybernetics: or Control and Communication in the Animal and the Machine.
- [4]House of Commons Treasury Committee. (2008). The run on the Rock. Fifth Report of Session 2007–08, Volume I. HC 56–I.
- [5]Meadows, D. H. (2008). Thinking in Systems: A Primer. Chelsea Green Publishing.
Suggest an edit· Updated 2026-10-02