A feedback loop occurs when the state or output of a system influences the process that determines its future state or output.

In an open-loop process, action proceeds without using information about the resulting output. In a closed-loop process, information about the output is fed back and can change subsequent action.

Negative feedback opposes deviations

In negative feedback, a change in the output produces a response that tends to oppose that change.

A common control architecture compares an output $y$ with a desired or reference value $r$. In a simple case where the output is fed back directly without rescaling,

$$e=r-y,$$

where $e$ is the deviation or error. The system acts on that deviation, and its action changes $y$.

Conceptually,

reference or desired condition
          ↓ compare
current output → error → corrective action
      ↑                    ↓
      └────── feedback ────┘

Negative feedback can reduce sensitivity to disturbances and uncertain component behavior, but it does not guarantee perfect regulation. Delay or overly strong corrective action can produce overshoot or oscillation.

Positive feedback reinforces change

In positive feedback, a change produces an effect that reinforces the original change.

output rises → action drives output still higher

Positive feedback can create rapid transitions, switching or self-amplification. Without another limiting mechanism, it does not generally stabilize a variable near a reference value.

Feedback is an architecture, not a domain-specific mechanism

The same logic appears in many systems:

• an engineered controller can adjust an actuator based on measured output;
• a biological regulatory system can change an effector response according to a sensed internal variable;
• an amplifier can feed part of its output back to its input circuitry.

The physical mechanisms differ, but the reusable idea is the same: the consequence of earlier output is routed back to influence later behavior.