Repair is frequently measured by restored functionality, even when the arrangement responsible for recurrent failure remains unchanged. This essay proposes RepairFeedback as a distinction between local correction and the consequential return of failure-derived knowledge to design, procurement, governance and training. The argument stages a dialogue between Wiener's feedback mechanisms, Bertalanffy's general systems perspective, Bateson's levels of learning and von Foerster's second-order cybernetics. These approaches illuminate different dimensions of institutional adaptation yet cannot be treated as interchangeable. Through a proposed comparative study of recurring defects in public facilities, the essay considers how evidence generated by maintenance workers and users might revise the decisions made by distant authorities. It also examines the danger that feedback systems transform repair into surveillance or continuous optimisation of an unjust structure. A repair becomes epistemically consequential when it changes an upstream decision, redistributes authority to identify failure, and allows subsequent recurrence to be measured. RepairFeedback names a traceable transformation rather than a celebration of resilience.
The Broken Thing and the Unchanged Institution
A repaired streetlight can illuminate a road again without changing the procurement decision that caused its repeated failure. A corrected software field can admit one previously excluded applicant while leaving the form's exclusionary categories intact. A museum may repeatedly restore an artwork with materials that reproduce the very deterioration requiring conservation. These cases reveal a gap between the repair of an object and the repair of the arrangement that governs it. Most maintenance accounting is oriented toward the first: time to restoration, replacement cost, service interruption and compliance. Those measures are necessary but incomplete if failures recur through choices made elsewhere. Socioplastics introduces RepairFeedback to distinguish an intervention that restores local function from a return path through which knowledge generated at the point of failure modifies upstream design, procurement, governance or instruction. The operator does not name feedback in general. Its distinctive question is whether the repair taught the institution something that it could not have learned from normal operation, and whether that knowledge acquired enough authority to change a decision.
This distinction recasts the temporal relation between designing and maintaining. Design is conventionally understood as the phase of anticipation; maintenance as the phase in which a finished system is kept operational. Yet workers who repair a system encounter materials under stress, uses outside the specification and conditions that modelled expectations failed to predict. Their experience should not be romanticised as automatically more truthful than engineering analysis, but its epistemic position is distinctive. It contains evidence produced by contact with breakdown. A functioning RepairFeedback loop must therefore connect observations at the site of failure to the locations where decisions are authorised, and then return revised specifications to the site. If an incident report is filed but never read, the loop is incomplete. If an executive changes a policy without documenting its effects, the loop remains untested. The productive unit is the circuit of evidence, decision, alteration and renewed observation, not the number of defects repaired.
Wiener: Feedback Is Not Yet Learning
Norbert Wiener's cybernetics established an influential vocabulary for control and communication in animals and machines. In a feedback arrangement, information about a system's output can be used to alter subsequent behaviour. The comparison with a thermostat is instructive precisely because it exposes a limitation. The thermostat corrects deviations from a desired temperature; it does not ordinarily deliberate over the social distribution of comfort or ask why that set point was selected. An institution can similarly improve its rate of repair while preserving a defective purpose. Faster restoration may represent considerable practical achievement and still leave the upstream specification unchanged. RepairFeedback should therefore not identify every negative-feedback mechanism as learning. A loop that stabilises an existing target differs from a process in which evidence from breakdown changes the target, the criteria of performance or the distribution of responsibility.
Wiener also complicates the tendency to treat information as neutral. Any feedback system must select what will be registered as a deviation. A public transport authority might collect delays but not the burdens imposed by overcrowded waiting areas; an urban greening programme might count failed plantings but not the labour required to keep surviving trees alive. Such choices belong to the architecture of observation. To make repair consequential, an inquiry should record not only whether a defect was corrected but what kinds of failure the institution is capable of detecting. It may become necessary to change the sensor, the category or the complaint procedure before an upstream redesign can occur. Here RepairFeedback touches FormatAdmission, but the operators should remain distinct: the latter examines what is allowed to enter as evidence, while the former follows the movement from evidence of failure to revised action. Together they expose how institutions can appear responsive while keeping their criteria of responsiveness untouched.
Bertalanffy: The Trouble with the Isolated Component
Ludwig von Bertalanffy's general systems theory contests explanations that assume a complex whole can be understood through isolated components alone. This principle is useful for interpreting recurrent failures whose causal conditions lie outside the damaged object. A ceiling may repeatedly develop moisture stains because drainage, thermal bridges, ventilation practice and institutional maintenance schedules interact. Replacing the stained panel repeatedly may improve appearance while masking the relations that reproduce the problem. RepairFeedback turns attention from the damaged element toward the configuration in which failure occurs. Its application requires tracing relations across physical systems and organisational divisions: what appears to the maintenance contractor as a component issue may originate in a design standard, budget cycle or management contract.
Yet a systems perspective also risks making boundaries too elastic. If every failure belongs to the whole system, no specific actor remains responsible for correcting it. Invoking complexity can become an alibi for diffuse accountability. Socioplastics therefore needs a distinction between contextual explanation and consequential intervention. The study must identify a tractable return path: a changed material specification, a different inspection interval, an amended contract or a revised public standard. The entire system need not be redesigned for feedback to count, but an identifiable upstream decision must change. This insistence also allows comparison across cases. Two institutions may face comparable failures while responding differently because one authorises technicians to recommend design modifications and the other restricts them to replacement orders. The question is not which system is more complex, but which arrangement permits operational knowledge to cross institutional boundaries and acquire decision-making force.
Bateson: Changing the Rule of Correction
Gregory Bateson's account of learning offers a more demanding interpretation of feedback. A system may learn to correct an error within fixed alternatives, or it may revise the context in which those alternatives have meaning. RepairFeedback draws on this difference without claiming an exact correspondence with Bateson's formal distinctions among levels of learning. A maintenance unit that replaces a broken hinge faster has improved its local procedure. A housing authority that asks why the hinge fails under predictable patterns of use, revises a procurement specification and changes how residents can report problems has altered a larger framework of practice. The distinction is practical but epistemologically significant. One operation corrects a deviation; the other revises the conditions by which deviations are defined and anticipated.
Bateson also makes learning inseparable from context and communication. An incident report is not a transparent container of objective facts. The language available to a worker, the urgency of the breakdown, the contractual relationship to management and the possibility of sanctions all shape what can be reported. A technician may know that repeated failures reflect a poor material choice yet classify them as user damage because the maintenance system offers no other admissible category. A feedback circuit can therefore fail before the evidence reaches a decision-maker. Its success depends on how contexts of communication change. RepairFeedback becomes intellectually useful when it identifies a specific passage across these contexts and can document a change of interpretation. It fails when the institution celebrates learning merely because more incident data have accumulated, while the categories and incentives governing their production remain identical.
Von Foerster: Who Observes the Failure?
Heinz von Foerster's second-order cybernetics sharpens this inquiry by refusing the fiction of an observer fully external to the system being described. The analyst who builds a performance dashboard determines what will be counted as failure; the institution that receives repair reports is also one of the forces determining how those reports are written. In an urban school, for instance, repeated overheating may be classified as a mechanical fault, a consequence of extreme weather, a teaching disruption or an inequity among classrooms. Each description carries different implications for responsibility. A feedback system that records the temperature but cannot register differences in authority among the observers may produce more data without producing better judgment.
A second-order account does not abolish material evidence. A broken ventilation unit remains broken regardless of the vocabulary chosen. It asks instead how observational positions shape the operational meaning of that evidence. RepairFeedback should therefore include the observer within its inquiry: who names the defect, who translates it into a maintenance category, who receives the report and who can decide that the category itself was wrong? Participatory repair might improve this process if those affected can alter reporting procedures, but participation can also create a new burden of unpaid monitoring. The challenge is to design an accountable circuit rather than simply invite more testimony. The repair must become legible across unequal positions without reducing the differences that made its diagnosis possible. The institution learns only when it can revise not just the object, but something in the way it observes and authorises its own correction.
A Comparative Test in Public Buildings
A testable research design could compare recurrent faults in two municipal facilities of similar age and programme over a defined period. The study would begin by identifying a bounded class of repeated incidents, such as water intrusion, overheating or damaged access equipment. It would map four stages: the first observation, the immediate repair, the transmission of evidence to a responsible authority and any revision of upstream rules. Documentary sources would include work orders, procurement records, complaints, drawings, meeting minutes and follow-up inspections. Interviews would help explain why some observations were transmitted while others were excluded. The relevant outcome would not simply be whether one building recorded fewer incidents. It would include the number and type of decisions actually changed, the time between observation and redesign, and whether the new arrangement redistributed burdens or created different failures.
The comparison would also need a negative case: a repair that appears successful because the visible defect disappears, but whose underlying cause continues to operate. Another counterexample is possible when a changed specification produces an unforeseen deterioration elsewhere. RepairFeedback is not a guarantee of wise institutional learning. It is a method for tracing whether a lesson travels back into the system and what happens when it does. Such an investigation could distinguish evidence of local skill, organisational responsiveness and transformations of standards. It might also expose a particularly troubling situation in which an institution converts repair data into worker surveillance, using every reported failure as evidence of individual inefficiency. The return path exists, but it travels toward punishment rather than improved infrastructure. The politics of the loop therefore belong to the definition of success, not to a moral appendix appended after the technical analysis.
The Right to Change the Design
The four interlocutors establish different conditions for an adequate theory of repair. Wiener makes the return of information technically intelligible; Bertalanffy resists the isolation of components; Bateson distinguishes correction from revision of learning contexts; von Foerster restores the observer to the processes of observation. RepairFeedback is strongest when it converts this intellectual constellation into a traceable question about authority: who can make the knowledge produced by failure change the arrangement that caused it? The answer varies between buildings, information systems, collections and public services. In each case, a repair can restore performance while concealing responsibility, or it can open a route through which practical knowledge transforms organisational decisions. The difference must be demonstrated, not inferred from a rhetoric of innovation.
Socioplastics itself would have to accept the same test. Identifying a weak conceptual relation or an erroneous bibliographic reference is not yet RepairFeedback. The operator becomes active when such findings change editorial rules, future classification practices or methods of verification, with the changes recorded and available for later criticism. The self-application is tempting but should not become evidence of universal validity. An archive, unlike a biological organism, does not automatically learn from what happens to it. People, institutions and material arrangements decide whether observations travel and whether they acquire authority. RepairFeedback makes that decision available for study. The political question at the end of the circuit is not whether a damaged system has returned to normal, but whether those who encountered its failure have gained any capacity to alter what normal will mean next time.
A second line of inquiry would examine the consequences of apparently successful repair for other institutional objectives. A revised public procurement rule might reduce component failure but increase dependence on one supplier, make local repair impossible or shift environmental costs toward production elsewhere. Such findings would not invalidate RepairFeedback; they would establish the need to document the scope and consequences of the new decision rather than counting any upstream alteration as success. The return of knowledge should itself remain open to later criticism. A repair culture that preserves incident histories can compare the reasons for consecutive revisions and identify when an earlier solution has become the cause of a new problem. This is particularly important in ecological interventions where the time scale of consequences exceeds the administrative cycle of a project. The strongest form of feedback is therefore not an organisation's claim that it has learned, but an accessible sequence in which another investigator can reconstruct the failure, the decision made in response and the evidence supporting or challenging its consequences.
Bateson, G. (1972) Steps to an Ecology of Mind. San Francisco: Chandler Publishing.
Bertalanffy, L. von (1968) General System Theory: Foundations, Development, Applications. New York: George Braziller.
Foerster, H. von (1981) Observing Systems. Seaside, CA: Intersystems Publications.
Wiener, N. (1948) Cybernetics: Or Control and Communication in the Animal and the Machine. Cambridge, MA: MIT Press.
Wiener, N. (1950) The Human Use of Human Beings: Cybernetics and Society. Boston: Houghton Mifflin.
Anto Lloveras is an architect and transdisciplinary researcher working across spatial production, cultural systems and situated inquiry. Through LAPIEZA-LAB and Socioplastics, he investigates feedback, maintenance and the institutional conditions under which knowledge survives and changes. His practice combines conceptual writing with architectural and curatorial experimentation.
