Feedback Loop

In 1988, Don Norman described what he called the action cycle in The Design of Everyday Things — a seven-stage model of how people interact with systems. Users form a goal, translate it into a plan, execute the action, and then evaluate whether the system responded as intended. That evaluation depends entirely on feedback. Without feedback, the user cannot know whether their action produced the intended effect, whether the system received it at all, or whether they need to try something different. The loop is broken.

Norman identified two types of feedback failure. The first is absence: the system produces no response, and the user is left in uncertainty. The second is delay: feedback arrives, but too late to be associated with the action that caused it. Research by Miller (1968) established that users perceive a 0.1-second response as instantaneous, a 1-second response as the boundary of uninterrupted thought, and anything beyond 10 seconds as requiring a progress indicator. These thresholds have not changed with faster hardware — they are perceptual constants.

“If feedback is not provided, the user is operating blindly — unable to learn, unable to correct, unable to confirm. The loop must close.”
— Don Norman, The Design of Everyday Things, 1988

Interaction feedback — the 100ms rule

Miller's (1968) research on response time tolerances established thresholds that remain the standard: under 100ms feels instantaneous, under 1 second feels connected, over 10 seconds breaks the sense of task continuity. The most common violation in modern product design is not loading time — it is the absence of any feedback at the moment of action.

Both submit buttons below trigger a 2-second processing delay — the same latency. The left provides no state feedback during that window. The right provides three sequential states: pressed, loading, confirmed. Click both to feel the difference.

The difference is entirely in what happens in the 2-second window between click and result. The left version provides no information — the user enters uncertainty that typically ends in a second click. The right version closes the loop three times. No uncertainty at any stage. The total time is identical. The felt experience is not.

Form validation — closing the loop at the moment of the mistake

Inline real-time validation and on-submit validation represent fundamentally different feedback loop lengths. On-submit validation produces a long loop: the user completes the entire form, submits, and receives all errors at once. Real-time validation produces a short loop: feedback arrives at the moment the mistake is made, before context is lost.

Baymard Institute's 2022 form UX research found that inline validation reduced overall form completion errors by 22% and decreased time-to-completion by 42% compared to on-submit validation.

Real-time validation closes the feedback loop at the field level rather than the form level. The user types an email, sees immediately that the format is incomplete, and corrects it before moving on. No context is lost.

Reinforcing loops — when feedback amplifies behaviour

A reinforcing feedback loop is one where the output of a system amplifies the input that produced it. In product design, this is the mechanism behind streaks, like counters, XP systems, and notification badges. Each works by making the result of an action visible in a way that motivates the next action.

Duolingo's internal data showed that users with streaks above 7 days churn at 5x lower rates than users with shorter streaks — not because the app got better, but because the feedback loop got stronger.

The same mechanism that makes Duolingo compelling makes social media platforms difficult to put down. The like counter, the follower count, the notification badge — each is a reinforcing loop. What differs is what behaviour it is amplifying and whether that behaviour serves the user's interests or primarily the platform's.