Systems Science 101 — Core Concepts
A feedback loop exists when the output of a system influences its own input. This circular causality is the defining feature of dynamic systems and the primary reason system behavior is so often counterintuitive.
Most human reasoning is linear: A causes B, B causes C, C causes D. Feedback thinking is circular: A causes B, which loops back and changes A. The loop changes everything.
Two Types of Feedback
All feedback loops are either reinforcing or balancing.
Reinforcing Loops (Positive Feedback)
A reinforcing loop amplifies change in the same direction. More produces more; less produces less. The name "positive feedback" refers to the direction of effect, not its value — positive feedback can drive a system toward catastrophe just as easily as toward growth.
Classic reinforcing loops:
- Compound interest — money earns interest, which adds to principal, which earns more interest
- Population growth — more people → more births → more people
- Word of mouth — more users → more referrals → more users
- Erosion — soil loss → less vegetation → more erosion
- Panic selling — falling prices → fear → more selling → lower prices
Reinforcing loops always produce exponential change — growth curves or collapse curves. Left unchecked, they run to limits.
Notation: Reinforcing loops are labeled R in causal diagrams. Every link in the loop has a consistent direction: all positive (same direction as cause), or an even number of negative links.
Balancing Loops (Negative Feedback)
A balancing loop resists change and seeks equilibrium. It compares a current state to a goal and acts to close the gap. All goal-seeking behavior is driven by balancing loops.
Classic balancing loops:
- Thermostat — actual temperature → gap from setpoint → heating/cooling → actual temperature
- Homeostasis — blood glucose → insulin response → blood glucose
- Inventory management — stock level → reorder trigger → deliveries → stock level
- Predator-prey — prey population → predator reproduction → predation → prey population
- Interest rate policy — inflation → rate adjustment → borrowing costs → economic activity → inflation
Balancing loops produce stability — or oscillation, if there is significant delay between sensing the gap and acting on it.
Notation: Balancing loops are labeled B in causal diagrams. They contain an odd number of negative links (links where the effect moves opposite to the cause).
Reading the Arrows
In a causal loop diagram, each arrow is labeled + or −:
- + (same direction): when the cause increases, the effect increases; when the cause decreases, the effect decreases
- − (opposite direction): when the cause increases, the effect decreases; when the cause decreases, the effect increases
Example: predator-prey
Prey population →(+)→ Predator births →(+)→ Predator population
Predator population →(+)→ Predation rate →(−)→ Prey population
Count the negative links around the loop: one (the predation rate → prey link). Odd number → balancing loop. This loop produces oscillation — classic predator-prey cycles.
Why Feedback Matters
Systems don't behave the way we push them. In a linear world, more effort produces more result. In a feedback world, the system responds to your intervention in ways that can amplify, dampen, or reverse your intended effect.
Delays make feedback dangerous. If there is a long lag between cause and effect, actors respond to outdated information. They overshoot and then overcorrect. The result is oscillation — and sometimes collapse.
High-leverage points are often in the feedback structure. Changing a parameter (a price, a rate, a quantity) rarely changes system behavior fundamentally. Changing the feedback structure — adding a new loop, removing a delay, shifting a goal — can change everything.
Feedback and Information
From an information systems perspective, feedback loops are signal processing architectures. Every loop carries information about a system's state and uses that information to modulate its own behavior.
This framing reveals several things that the standard control-systems view underemphasizes:
- What the system measures is as important as what it does with the measurement. A sensor that measures the wrong variable — or measures the right variable with too much delay — produces a feedback loop that regulates the wrong thing, or regulates too slowly to be effective.
- What the system ignores is equally important. Every feedback loop is selective. It responds to some signals and filters out others. The design of this selectivity determines what the system is actually managing.
- Attractor states are defined by feedback structure. A balancing loop with a fixed goal defines a fixed-point attractor — the system will return to that point after perturbation. Multiple interacting loops define a richer attractor landscape: stable points, cycles, and in complex cases, strange attractors. The feedback structure is the attractor structure.
This is why understanding feedback loops is not just about tracing causal arrows — it is about understanding what information the system is using, what it is optimizing for, and where it will naturally settle.
Feedback in Practice
Recognizing feedback loops in real systems:
- Identify the key stocks — what is accumulating or depleting?
- Ask what drives the flows — what controls how fast the stock changes?
- Trace the influence back — does the stock itself influence the flows? Through what path?
- Identify the goal (for balancing loops) — what state is the loop trying to achieve?
- Find the delays — how much lag separates cause from effect?
A system with multiple interacting feedback loops — some reinforcing, some balancing, with varying delays — can produce any behavior: growth, oscillation, collapse, S-curves, or chaotic fluctuation. The behavior is in the structure.
Part of the Systems Science 101 reference. See also: Emergence · What Is Systems Science? Deep dive: Systems Dynamics course