Definition
The principle that a case a representation leaves out may be evidence about the category, not merely noise around it. It is the epistemological ground of institutional fallibilism. A complaint can be knowledge. Repeated exception handling can reveal more about a system than its official performance metrics, because the exclusions are where the representation's boundaries become visible.
It matters because it is why dissent and standing are epistemic and not only political. Knowledge stays defeasible only if evidence the existing representation did not know how to recognize can still arrive — and it arrives through the people who carry it. That makes fallibilism material: a system is not meaningfully open to correction if the people who know it is wrong must exhaust themselves to prove it, which is the condition corrigibility rent prices.
Tests that show it is in place
- Cases outside the model's support can enter as challenges rather than being discarded as noise.
- Exception patterns feed category review, not only case resolution.
- The cost of raising an excluded case is measured against the challenger's capacity to bear it.
Evidence to keep
- Artifact: An exceptions register where out-of-support cases are recorded as potential category evidence.
- Behavior: An excluded case that produced a category or rule review rather than a case-level fix.
- Metric: Share of exception classes whose pattern reached a category or rule review.
References
- Exception learning · Theory
- Laws · Doctrine