feedback_auto_deep_dive_via_framework_rediscovery

Rule: When a scorecard’s top options are close AND share a structural limitation, don’t re-weight criteria. Instead, ask: “what recently-locked framework haven’t I applied to this decision?” Then apply it to produce a dominant refinement option.

Two trigger modes (both validated 2026-04-24):

1. Narrow-margin-triggered (<5pp threshold; original rule)
2. Elective framework-application (margin above 5pp but a locked framework would produce a structurally-dominant option)

Why: This pattern emerged through 3 auto-deep-dive wins in morning 2026-04-24, then reinforced across 5 more wins in afternoon/evening — not coincidence, but a reliable method. In all 8 cases, applying an unused locked framework produced a refinement not in the original option set.

Session 2026-04-24 — 8 confirmed wins

#	Adjudication	Narrow margin	Missing framework	Refinement produced	Dominant margin
1	Q4 SD1 scope package	α/γ at 5pp	feedback_introduction_module_rule (AC-1 two-rule)	δ Infra-abstract + intro-rule distribution	11.0pp
2	T-NEW-02 SD3 polygamous scope	α/ε at 5pp	A-23 vi-6 OMG Commons Classifiers lock	α’ Household + householdRegime classifier	7.2pp
3	T-NEW-CL-05 parent SHACL discipline	D/F at 1.2pp	Three-category structural analysis	H 2-amendments + A-21 CI extension + guide	3.4pp
4	T-NEW-CL-06 parent public-policy contracts	D/F at 2.0pp	A-23 vi-6 (2nd application) — MaturityScheme	J 4-split with MaturityScheme refinement	0.6pp
5	T-NEW-CL-02 SD3 intervals-primary	α/ε close	PROV-O + intervals-primary encoding	A* intervals-primary + ROBOT-derived prov:wasRevisionOf**	5/5 normalization (99.6%)
6	T-NEW-08 SD1 error-code seeds	ε/γ 8.5pp ELECTIVE	A-51 + A-23 vi-6 + v1-regression-preservation	ε’ 5-category + 8-v1-leaves hybrid	6.0pp
7	T-NEW-01 parent executed-state	C/D at 1.6pp	A-23 vi-6 + AC-1 + intro + v1-regression	C’ Hybrid + FailureReasonScheme	5.0pp
8	T-014 parent SDK coverage	A baseline 78.9% ELECTIVE	A-55 MaturityScheme extension to SDK layer	G 5-lang + tier-aware governance	6.6pp

Pattern diversification: wins 1-5 were all narrow-margin-triggered. Wins 6 and 8 were ELECTIVE (margin above 5pp; surfaced because applying available framework produced dominant refinement). Win 7 straddles — 1.6pp narrow margin with 4-framework convergence.

How to apply (two modes)

Mode A: narrow-margin-triggered

1. Detect the signal. Top options <5pp apart. Shared structural limitation.
2. Inventory unused frameworks. Walk through recently-locked amendments + feedback memories.
3. Candidate screen. Does any framework produce an option not in original set?
4. Score the candidate. Expect a dominant margin.
5. If multiple candidates produce similar scores: apply combined refinement.

Mode B: elective framework-application (new 2026-04-24 evening)

1. Detect the signal. Scorecard has a clear baseline winner (above threshold), BUT an unused locked framework could produce a structurally-richer option.
2. Test the framework-application. Quick mental check: would applying X framework produce an option that dominates the current winner?
3. Surface the refinement if yes. Don’t suppress because “margin is already clear” — session convention is to surface and let Rich pick.
4. Honest framing. If elective, say so — “this is not a narrow-margin rescue; it’s an available framework-application.”

Signals that the pattern applies

• Narrow margin feels arbitrary — scorecard splitting hairs not discriminating real choices.
• Both top options criticised from same angle.
• A recent memory or amendment feels “relevant but not consulted.”
• Rich asks “can you study these options more carefully?” — strong signal.
• Elective signal: gut-check “this option is fine, but I know X framework would make it better.”

Signals the pattern does NOT apply

• Top options differ structurally on fundamentals (different trade-offs, not shared limitations).
• Narrow margin reflects genuine value-preference ambiguity (e.g., reversibility vs visibility).
• No unused framework surfaces during inventory.
• Elective mode: no locked framework produces a materially-different option.

Meta-observation (2026-04-24 evening): framework compounding

The more locked frameworks the architecture accumulates, the more likely future adjudications will resolve via framework-rediscovery:

• Morning (wins 1-5): 5 wins across 8 tensions = 62.5% framework-triggered
• Evening (wins 6-8): 3 wins across 3 tensions = 100% framework-triggered
• Full day (wins 1-8): 8 wins across 11 tensions = 72.7% framework-triggered

Late-session decisions compress to “which locked frameworks converge here?” rather than “what’s the right answer?” Architectural maturity accelerates decision-making.

Most-reused session frameworks (2026-04-24)

Framework	Applications	Domains
A-23 vi-6 SKOS classifier scheme	7	HouseholdRegime / Maturity / LegislativeStatus / CitationType / ErrorCode / FailureReason / CommorientesPresumption
A-55 MaturityScheme	4	primitives (origin) / SDK tiers / SDK promotion criteria / CI severity
Introduction-module rule (AC-1 two-rule)	3	Q4 SD1 / T-NEW-01 parent / T-NEW-01 SD1
A-51 crate-prefix naming	2	Error-codes (extended to SKOS Concepts) + failure-reason partner crates
PROV-O machinery	2	ExecutedEstate+BequestResolution (Activity) / DistributionContext (Entity)
v1 regression-preservation	3	T-NEW-08 SD1 / T-NEW-01 parent / T-014 SD2
Law-vs-practice split	3	T-NEW-08 / T-NEW-01 / T-014 MCP

Relationship to existing memories

• feedback_automatic_deep_dive_when_options_feel_imperfect — general trigger rule; this memory specifies the method
• feedback_reframe_beats_reweight — principle this pattern implements
• feedback_scorecards_one_at_a_time_optimal_sequence — sequencing discipline; each locked SD makes next SD’s frameworks available for rediscovery

Cross-reference: State file inherit-v2-architecture-state.md v2.10 changelog notes auto-deep-dive count at 8 wins cumulative.