Correctability

Reality Contact requires two capacities, and either can fail while the other still works. The first is orientation: reading present position against a fixed reference as conditions change. It has its own page. This page is the home of the second. Correctability is whether reality can still correct the system: whether an accurate reading can reach the controls and change the course, and whether that path stays open as load accumulates. The pages are separate because the failures are separate. A system can see clearly and be uncorrectable, and that combination is where the most dangerous failures in the record begin.

The frame to set aside first is self-repair. An institution does not become correctable by resolving to improve. It becomes correctable when its structure leaves a path open from evidence to consequence, so that when the world contradicts it, the contradiction can land somewhere and change something. Correctability is the capacity to be corrected, and the passive voice is deliberate: the system is the object of the verb. Reality is the subject. Everything else on this page describes the conditions that arrangement depends on.

Correct descends from the Latin corrigere: to set straight. A sheet metal shop still uses the idea daily. Before the first cut, a layout line goes down on the metal, and every operation afterward is measured against it. When the work wanders, the line is what the work is guided back to; it is the reference that makes back a direction.

The shop's most automated machine obeys the same rule. Before the laser cuts, it finds the edge of the actual sheet on the table, not the sheet its program assumes, and corrects its coordinates to what it finds. The intricacy of the cut is earned by that moment of consulting the material before trusting the model. Correction, in the sense this corpus keeps, is exactly that: a return to a reference that never moved.

Take the two capacities apart and three conditions are possible. A system that can neither read its position nor be corrected has lost both components of Reality Contact at once; nothing anchors it to the world it is moving through. A system that can do both is answerable: it will still be wrong at times, and the wrongness will cost it something and change it. Between them is the quiet condition, the one that matters most here: the system reads accurately and cannot be corrected. The instruments are true. The reports are right. Somewhere between the reading and the controls the path is closed, and the system continues its course with a clear picture of the terrain it is failing to respond to. The corpus documents cases in which exactly this happened: accurate readings, produced in time, that changed nothing.

What keeps the path open from an accurate reading to a changed course is itself under active study here, and the map is not finished. Among the conditions the corpus has identified as bearing on it: the reading needs standing, admission into the decision in time to matter. Authority needs to be connected to evidence and consequence, so that acting on the reading is a specific person's job with a specific person's name on it; Decision Integrity Under Pressure examines what load does to that connection. And the record needs to survive with its meaning intact, because a correction the institution cannot remember making is difficult to keep; Memory Continuity covers that. When conditions like these degrade, no one repeals correctability. The path narrows until an accurate reading no longer fits through it. The loss is not a verdict. Degraded conditions can be rebuilt, and the corpus's restoration standards specify that work; what cannot be assumed is time, because the recoverable window closes without announcing itself.

Every correction makes a claim: that it reached the structure which produced the failure. Recurrence is one of the sharpest tests that claim can face. When the same class of failure comes back under comparable conditions, after the investigation, the report, and the reforms, that return is evidence that the last correction failed to reach, or failed to persist, and the interval between returns says something about where it stopped. The absence of recurrence proves less than it seems to: conditions change, exposure ends, activities stop, and luck intervenes. Recurrence and Correction tracks the record of returns through the documented cases.

Correctability concerns whether a system can still be guided back. Toward what is a different question. What a system should correct toward is a commitment, made and defended on grounds the science does not supply; the science describes the conditions under which any commitment, whatever its content, stays answerable to what actually happens. That boundary is deliberate, and Restorative Realism describes what lives on the other side of it.

David D. Woods · The stretch

David D. Woods's resilience engineering, particularly his Theory of Graceful Extensibility (2018), asks whether an adaptive system can extend its capacity for maneuver when surprise pushes it toward the limits of what it was built to do.

Correctability asks a neighboring question: whether contradiction from the world can still acquire force over what the system decides. The distinction matters because a system can extend its ability to continue doing something without preserving the ability of reality to tell it that it should be doing something different. Woods's account of limited perspectives, and of adaptive units improving through contrasting perspectives, may also illuminate part of how corrective pathways form. The relationship between the two bodies of work is therefore one to investigate, not presume.

Nancy Leveson · The control structure

Nancy Leveson's STAMP treats safety in complex sociotechnical systems as a control problem, examining how constraints, feedback, process models, and control actions succeed or fail across hierarchical control structures, from hardware through operators, management, and regulators. Those structures include the means of revising their own models and constraints, and STAMP's causal scenarios describe many of the ways feedback and revision degrade in practice.

Correctability approaches the same territory from a different direction: it studies, as a property in its own right and for systems of any purpose, whether contradiction from the world can still acquire force over what a system decides. STAMP evaluates control structures against defined losses; Correctability asks whether reality can still reach the decision at all, wherever decisions produce consequences in the world.