MOC: Schema change and restructuring — how knowledge structures update, and when the vault should rewire itself

The cognitive-science thread harvested from "when and how should SeekVault deliberately rewire itself" — the counterpart to the gap-detection cluster (moc-knowledge-gap-detection finds what's missing; this cluster is about what happens when a structure has to change). Spine idea, recurring across every note: prediction error is the shared trigger, and the friction of a misfit is the signal you must not smooth away.

The taxonomy (what the modes are)

• claim-schema-change-four-mode-taxonomy — accretion / tuning / restructuring / schema formation, each PE-triggered (Rumelhart-Norman 1978 → SEEK model 2026); plus the caveat that "restructuring" changed meaning.
• claim-kgcl-defers-merge-split-topology-changes — the same operations in a knowledge-graph change language: node/edge edits are built, merge/split (the topology-reorganizing pair) are explicitly future work.

The trigger (what starts restructuring)

• claim-cognitive-conflict-prerequisite-for-restructuring — conflict is a prerequisite; congruent information cannot trigger it.
• claim-p3-indexes-prediction-error-schema-updating — the P3 potential tracks PE magnitude predicting schema update; it does not mark a tuning/restructuring distinction (premise correction).
• claim-bjork-desirable-difficulties-not-prediction-error — the guardrail: Bjork 1994 is not a prediction-error source (verified negative); don't rest the PE spine on a mis-cited anchor.

The architecture (is change one process or several)

• claim-insight-conflict-and-restructuring-distinct-networks — conflict and restructuring are partially-overlapping-but-distinct brain networks: neural evidence the two operations are separable.

The obstacle (why restructuring fails)

• claim-expedient-knowledge-blocks-restructuring — a good-enough strategy learned early blocks later restructuring; accuracy makes bad structure sticky.

The cross-domain frames (organizational punctuated equilibrium; cognitive-linguistic frame theory)

Domains outside the cognitive-science core that supply the same equilibrium/revolution shape — a durable deep structure that resists ordinary pressure. Organizational-change theory contributes both notes below (also a rival account of the trigger); both are seedling on [unverified-mechanism] (primary text access-walled): question-verify-organizational-punctuated-equilibrium-primaries.

• claim-gersick-1991-punctuated-equilibrium-deep-structure — long equilibrium periods within a durable "deep structure," interrupted by brief revolutionary reorganizations of that structure.
• claim-tushman-romanelli-1985-reorientation-exogenous-trigger — the trigger is partly exogenous (environmental jolts, executive succession), not accumulated internal strain — a direct complication to the endogenous-threshold story above. The vault-facing hypothesis it spawns: question-new-session-onset-as-rewiring-trigger.
• claim-lakoff-frames-neurally-entrenched-resist-reframing — a third domain supplying the same shape: cognitive linguistics. Lakoff argues political frames are neurally entrenched structures that strengthen with use (even under negation) and reframe only through deliberate repeated effort. seedling on [unverified-mechanism] — the synaptic-entrenchment mechanism is his own framing, not adjudicated neuroscience: question-verify-lakoff-neural-frame-entrenchment-neuroscience. (Same intellectual parent, Metaphors We Live By, that gave software engineering "technical debt": claim-cunningham-technical-debt-from-lakoff-metaphor-theory.)

The bridge to backpropagation's origins (the author is the hinge)

• observation-rumelhart-person-bridge-schema-theory-to-backprop — David Rumelhart authored both the 1978 tuning/restructuring taxonomy above (claim-schema-change-four-mode-taxonomy) and, as the "R" in RHW 1986, the backprop revival in moc-backpropagation-origins. One trajectory; a person-bridge invisible to embedding retrieval because no note named the hinge. seedling — the stronger "schema-tuning is the ancestor of weight-tuning" reading is left open at question-verify-rumelhart-schema-tuning-to-gradient-descent-through-line.

The vault's own open decision

• question-consolidation-pass-vs-revisit-protocol — targeted hooks+lint vs a standing global consolidation pass. First-lint evidence (n=63) leans against a global pass; re-measure as the note count and retrieval load grow.

The through-line for Seek

The literature backs treating "spin up a new cluster" and "reorganize an existing one" as distinct operations — but NOT the vault's rule of thumb that one needs a bigger trigger than the other (that's still a guess; claim-schema-change-four-mode-taxonomy). And the sharpest warning is claim-expedient-knowledge-blocks-restructuring: the better structure has to arrive early or accumulated habit blocks it — the north star's argument for wiring the embedding layer at threshold, not after the workflow calcifies.