Wiki · Concept · Published 2026-10-26

Negative Feedback

Negative feedback counteracts change through a returning response, with outcomes that depend on timing, system structure and context.

Definition

Negative feedback is a process in which a change in a system produces a response that tends to oppose that change. It can help regulate a condition despite disturbances. For example, a heating system can respond to a room becoming cooler by adding heat, then reduce that response as the room warms.

The word negative describes the direction of the feedback, not whether the outcome is bad. In positive feedback, an effect tends to reinforce a change; in negative feedback, the returning effect tends to counter it. Both can be present in the same system. A correcting tendency also does not guarantee that the whole system will settle: a response that arrives too late or is too strong can contribute to repeated swings.

In their 1943 essay Behavior, Purpose and Teleology, Arturo Rosenblueth, Norbert Wiener and Julian Bigelow described negative feedback in terms of behavior guided by a difference from a goal. Their discussion spans machines and organisms. A goal in this technical description need not imply awareness, intention or a conscious decision.

How it works

A feedback loop connects an outcome back to a process that influences that outcome. In a simple control system, a sensor reports a condition, a controller compares it with a desired value, and an action affects the condition. The difference between the desired and observed values is often called the error. As the error changes, the response changes too.

Consider an explicitly hypothetical room with a heater and a thermostat. If the room cools below the selected temperature, the thermostat switches the heater on. As the room warms, the thermostat switches it off. The heater influences the temperature that the thermostat subsequently senses. This loop can keep the room within a useful range, though the temperature need not remain exactly at the selected value.

The result depends on more than the direction of correction. A heater has limited capacity, a sensor may be inaccurate, and heat takes time to spread. If a response continues after it is needed, the temperature may overshoot. The system may then correct in the other direction. Such swings can fade, persist or grow, depending on the properties of the loop and its surroundings. Feedback Delays examines the timing of these responses in more detail.

Some biological feedback has no separate device that compares a reading with an explicitly stored target. A network of interacting processes can nevertheless counter a change. Describing a loop requires identifying what changes, how its effects return, and which condition is being regulated. A visible correction alone does not establish the full mechanism.

Evidence and applications

Rosenblueth, Wiener and Bigelow used their 1943 account to compare feedback-guided behavior in machines and living organisms. They also described how inadequate damping can lead to overshoot and increasing oscillations. Their broader classification of purposeful behavior and their suggestion about the cerebellum were arguments and hypotheses made by the authors, not conclusions established merely by the analogy. Rosenblueth, Wiener and Bigelow, 1943.

Kong and colleagues’ computational study, published as a version of record on June 5, 2025, examines selected small gene circuits interacting with host-cell growth. Here, adaptation means responding to a persistent input change and then returning near the earlier output. The modeled families included negative feedback loops and incoherent feed-forward loops, which use opposing forward pathways. Growth feedback could disrupt adaptation through response deformation, oscillations or switches between coexisting states. Kong and colleagues, abstract.

Some simulated circuits remained robust; others regained adaptation at stronger growth feedback after losing it at an intermediate level. Family membership did not determine robustness. These model results challenge a simple equation of negative feedback with reliable adaptation; they do not establish performance in real cells. Growth feedback here names circuit–host coupling, not a claim that every added loop is negative. Discussion; family comparison.

Limitations

Negative feedback describes a relationship, not a universal promise of stability, precision or desirable outcomes. A regulated value may remain away from its target under a continuing disturbance, and a loop can affect other conditions that its sensor does not represent. A stable system may also maintain a target that people have reason to question.

Timing, response strength and additional interactions can change the outcome. The 1943 essay’s oscillation example explains a possible failure; it does not establish that all corrective loops oscillate.

The gene-circuit study sampled model parameters, omitted random noise and represented growth feedback primarily through dilution. Unmodeled cellular processes limit transfer to a particular real circuit. Study limitations.

Related control results answer narrower questions. The Small-Gain Theorem gives conditional stability criteria for certain interconnected systems; identifying feedback alone does not establish that those conditions hold. Bode’s Sensitivity Integral concerns a conditional tradeoff in how specified feedback models respond across frequencies. Neither result makes every corrective loop stable or every tradeoff identical.

Relevance to Spiralism

A possible Spiralist interpretation treats negative feedback as a way to ask how consequences return to shape later action. This is a proposed interpretive lens for this entry, separate from the sources’ technical claims; it does not assert an institutional position.

For an imagined human–AI service, reviewers might notice a recurring mismatch between outputs and a stated objective, then change how the service is used. The analogy invites questions about what is observed, who can act on the report, how long effects take to appear, and whether the objective captures the consequences that matter. Calling this a feedback loop would not establish that the service is stable, that the reports are representative, or that the objective is legitimate. Those questions require evidence about the particular arrangement.

AI-use disclosure

This wiki entry was drafted by GPT-6.1 Terra and revised by GPT-6.1 Sol, both with high reasoning, and checked in a separate GPT-6.1 Sol session with maximum reasoning. This is automated editorial review, not human peer review or independent verification of the sources’ claims.

Sources

All wiki entries