Feedback loop
A three-step cycle at the heart of learning a skill: the learner produces work, compares it against some reference, and acts on the difference. The loop only does its job when the learner can act on the comparison — feedback that arrives but cannot be used leaves the loop open and changes nothing. How well it works depends on what the feedback addresses, how soon it arrives, and whether the learner has room to close the gap.
The cybernetic structure that the skill-formation literature treats as the active component of any skill-building method. The minimal form is three steps: the learner produces work, the work is compared against a reference, and the comparison is acted on. The reference may be an external criterion (a worked-out solution, a rubric, an expert demonstration), an internal one (the learner's own intention for the work), or both. The closed loop is the one where the learner can act on the comparison; the open loop — feedback given but not actionable — is what most institutional grading produces. The loop's effectiveness depends on what level it addresses (task, process, self-regulation, or self), how soon after the work it arrives, and whether it is paired with structural pressure toward closing the gap.
Etymology§
The cybernetic vocabulary of feedback in this technical sense entered learning research through Norbert Wiener's Cybernetics (1948) and the wider mid-century engineering-and- biology effort to describe self-regulating systems. The term travelled into educational psychology gradually through the 1960s and 1970s. The contemporary anchor in the topic's literature is Sadler's 1989 three- condition account of formative assessment (visible criteria, comparison, action on the gap) and Hattie and Helen Timperley's 2007 Review of Educational Research paper, which organised the empirical literature around what feedback addresses and what conditions determine whether it works.
The loop's three elements are not all equally available in any given instructional context. Producing the work is generally easy to arrange; the structural difficulty is in the comparison step and the action step. The comparison requires a reference the learner can read against — and the literature is consistent that the learner has to understand the reference, not merely be exposed to it. A rubric pinned to the wall does not make formative assessment work; the rubric has to be operationalised into instances the learner can read their own work against. The action step requires that the learner has time, permission, and the cognitive resources to do something with the comparison — which in most institutional timetables they do not.
The four-level model from Hattie and Timperley is the standard decomposition of what feedback addresses. Feedback at the task level says what was correct or incorrect about the specific work produced. Feedback at the process level says something about the method used to produce the work — what strategy was applied, what alternative might have been better. Feedback at the self-regulation level addresses the learner's monitoring of their own work — how they knew when to seek help, when to revise, when to stop. Feedback at the self level addresses the person rather than the work (you are good at this, you didn't try). The first three help; the fourth, on the empirical evidence the 2007 paper synthesised, hurts.
The empirical complication is Butler's 1988 result: even feedback given at the task or process level loses its formative effect when a score is attached. The score acts as a self-level summary that occupies the attention slot the formative content would have used. The implication that the literature has been working with for thirty years is that grades and formative feedback compete for the same cognitive resource in the learner, and that running both together typically collapses to the score.
The loop's relationship to deliberate practice is direct: deliberate practice is a feedback loop, with the feedback condition specified structurally rather than left to whatever a teacher happens to provide. Ericsson's four conditions (narrow goal, immediate feedback, repetition with adjustment, full attention) are an operational decomposition of what makes a feedback loop produce skill gain rather than mere repetition. The loop's relationship to mastery learning is the same: mastery learning is a structurally enforced feedback loop, with the institutional commitment that the loop runs until the criterion is reached rather than being terminated by a calendar.
Writing is the standard hard case for the loop. The writer can re-read their own work, but the re-reading runs into the same cognitive blind spots that produced it; a peer or editor supplies external feedback only at low frequency. Some of the contemporary AI-assisted writing tools function as feedback sources at higher frequency than any human reader can, which is part of what the Mollick strand of the literature is currently working through.
Discussed in§
- Inside the Black Box: Raising Standards Through Classroom Assessment
- Enhancing and Undermining Intrinsic Motivation: The Effects of Task-Involving and Ego-Involving Evaluation on Interest and Performance
- The Power of Feedback
- Co-Intelligence: Living and Working with AI
- Formative Assessment and the Design of Instructional Systems