Learn it → Apply it → Save as Skill
Every system talks back to itself. A feedback loop happens when a system's output becomes its own next input, either amplifying whatever is already happening or pulling the system back toward balance. Most surprises in business, habits, and relationships are loops you didn't notice running.
The Dr. Wallace's Favorite Mental Models is designed to turn useful ways of thinking into reusable tools. Each model gives you a fast visual overview, deeper interactive learning, practical application exercises and an AI-ready Markdown skill you can save for future use.
Instead of merely reading about mental models, you can gradually build your own AI skills library: a collection of portable reasoning skills that can help ChatGPT, Claude, Grok and other AI assistants apply the mental models you value.
A feedback loop forms whenever what a system produces circles back and changes what it produces next. A savings account's interest becomes next month's balance. A product's user growth becomes next month's word of mouth. A thermostat's room temperature becomes the next signal that tells the furnace whether to run. Notice the loop, and you can often predict, or redesign, where the system is heading.
Important: "reinforcing" and "balancing" describe direction, not good or bad. A reinforcing loop can be a growth engine or a death spiral; a balancing loop can be a healthy home thermostat or a bureaucracy resisting a change it actually needs.
Most useful the moment a metric, habit, or relationship seems to be accelerating or stubbornly resisting change, before you assume a single cause.
The method traces the loop before you try to change it, so your intervention hits the actual mechanism.
Draw the chain: what output feeds back in as an input, and through what delay?
Decide if the loop is reinforcing (amplifies the current direction) or balancing (resists it, seeks a target).
Locate the weakest link: the slowest delay, the leakiest step, or the rule that sets the target.
Strengthen a reinforcing loop you want, dampen one you don't, or shift what a balancing loop is targeting.
Open a use case to see what the model changes.
Before crediting the marketing team alone, check for a reinforcing loop: more users generate more reviews or referrals, which drive more signups. If so, the growth compounds on its own, and the real lever is strengthening the referral step, not just spending more on ads.
Overloaded teams often ship worse work, which creates more rework, which eats the capacity meant for new work. That is a reinforcing loop pulling the team further from its estimates each sprint. Fixing the estimate without breaking the rework loop rarely holds.
A habit without a balancing loop, some form of tracking or feedback that shows you drifting off target, tends to erode once motivation fades. Adding a visible signal (a streak, a weigh-in, a weekly review) gives the habit something to correct against.
Markets usually run on balancing loops: a price rise cools demand, which pulls the price back down. But scarcity or status goods can flip into a reinforcing loop, a rising price signals desirability and pulls in more buyers, which explains bubbles and hype cycles.
Melting ice reduces the reflective surface that cooled the region, which warms it further and melts more ice. Recognizing this as a reinforcing loop, rather than a one-time event, changes how urgently and where an intervention needs to land.
One sharp comment invites a sharper reply, which invites a sharper one back, a reinforcing loop with no target. Naming it out loud, "we're escalating," can act as a deliberate balancing loop that interrupts the pattern before it runs further.
This is an illustrative example designed to teach the method.
Expand for the research and terminology behind the model, and how confident the field really is.
Mechanical feedback predates the theory: James Watt's centrifugal governor (1788) automatically throttled a steam engine using its own speed as the input, a working balancing loop built before anyone named the concept. The formal theory arrived with mathematician Norbert Wiener, whose 1948 book Cybernetics: Or Control and Communication in the Animal and the Machine named and generalized feedback as a control principle spanning machines, organisms, and organizations.
Jay Forrester, an MIT engineer, founded the field of system dynamics in the 1950s and applied feedback-loop modeling to factories and cities in works like Industrial Dynamics (1961). Donella Meadows, a member of that MIT circle, later wrote the widely cited 1999 paper "Leverage Points: Places to Intervene in a System" and the posthumously published Thinking in Systems: A Primer (2008), both standard references for applying loop thinking outside engineering.
Yes, "reinforcing loop" and "positive feedback" describe the same mechanism, and no, positive does not mean beneficial. In control theory, positive and negative describe the direction the loop pushes a variable, not whether the outcome is desirable. A reinforcing loop can just as easily describe a bank run as it can compounding savings.
System dynamics is a well-established engineering and management discipline with decades of peer-reviewed modeling work, but applying it to messy human systems, an economy, a company culture, involves real judgment calls about where the true boundaries of the loop sit. Meadows herself cautioned that intervening at the wrong leverage point, even with an accurate loop diagram, can fail or backfire; treat loop diagrams as a structured hypothesis to test, not a guaranteed prediction.
Not every trend is feedback. A one-time external shock, a holiday sales spike, a single viral post, can look like a reinforcing loop without any self-sustaining mechanism behind it.
Debt spirals, arms races, and outrage cycles are reinforcing loops too. The label only describes direction, never desirability.
Many loops have a lag between output and effect. Judging a loop by its first few data points, before the delay plays out, produces a badly wrong read.
Treating a symptom, adding more of the thing running short, instead of the loop's actual bottleneck, wastes effort and can even strengthen the problem loop.
Real systems usually run several loops at once, as in the worked example above. Fixing one in isolation can simply hand control to another loop you didn't map.
A ten-arrow causal map that nobody can act on is worse than a rough three-step loop that clarifies the one decision in front of you.
a metric is accelerating, decelerating, or oddly stable in a way a single cause can't explain, and the pattern has repeated more than once.
a problem traces cleanly to one identifiable event or defect with no ongoing self-reinforcement, such as a single server outage.
the system is genuinely novel and no one, including you, has enough observed data yet to trace a credible loop; guessing a mechanism from too little evidence can mislead more than it helps.
This browser tool structures your thinking. It does not send your information anywhere or perform AI analysis.
Save as Skill converts the model into portable Markdown instructions that teach an AI when and how to apply Feedback Loops, what questions to ask, what evidence to seek, what mistakes to avoid and how to structure its reasoning.
Save the skills you find useful and organize them into your own reusable library. The Markdown is intentionally model-neutral, so it can be used wherever an AI system accepts uploaded files, project knowledge, custom instructions or reusable text context. Platform capabilities differ, so this is a portable skill file rather than a claim of one-click installation.
Vicious and virtuous cycle are everyday names for a reinforcing loop, the vicious version pushing an unwanted outcome, the virtuous version pushing a wanted one. The systems-thinking vocabulary (reinforcing and balancing) is more precise because it describes the loop's structure, not just whether you're happy with where it's heading, which matters once a loop's direction changes.
Yes. A balancing loop that loses its correcting mechanism, a thermostat with a broken sensor, a market losing its price signal, can start behaving like a reinforcing loop with nothing to stop it. This shift is a common cause of systems that were stable for years suddenly running away.
A leverage point is a place in a system where a small change produces a large shift in behavior. Meadows ranked leverage points from weak (tweaking a numeric parameter, like a tax rate) to strong (changing the goal the whole system is organized around, or the mindset that produced the goal). Her key caution: people, including experts, tend to intervene at the weakest, easiest points and wonder why nothing changes.
Write down what data would exist if your loop hypothesis is wrong, before you check. If reinforcing growth is really driven by referrals, a period with no new invites should show growth stalling. If it doesn't stall, the loop you drew probably isn't the real mechanism.