Review Essex evidence-to-object fit and model-led follow-up

[happygeneralist]

Summary

Adds a public-safe analytical review of the first bounded Ofsted Essex evidence-to-object chain and an operational correction clarifying that the next ingestion test should use the existing assumptive journey analysis through the repository's model-led, non-linear ingestion route.

Changed files

• docs/ingestion-slices/ISSUE_087_ofsted_essex_evidence_to_object_fit_review.md
• docs/ingestion-slices/ISSUE_087_model_led_ingestion_operational_correction.md

Evidence-to-object finding

The evidence-to-PP_002 connection is directionally useful but only a partial fit.

The source supports the communication, support-visibility and unresolved-questions manifestation of navigation burden. It does not substantiate the whole broad pain point, its severity: high judgement, sustained coordination work, chasing behaviour, work/caring impacts or national recurrence.

The review therefore keeps all current object states unchanged and recommends that the evidence link be interpreted as bounded support rather than wholesale validation.

Journey-pattern finding

The report suggests a provisional lived-experience pattern around:

• maintaining agency and orientation during institutional waiting; or
• waiting without reliable orientation.

This is framed as a non-linear pattern involving waits, support visibility, communication reliability, compensating effort, escalation and effective-support branches. It is not proposed as a canonical journey stage or new object.

Operational correction

The original review preferred a contrasting direct lived-experience source as the next step. That is too restrictive and does not reflect the repository's existing non-linear ingestion model.

The second note records two supported routes:

Evidence-led
source -> evidence -> bounded interpretation -> object reconciliation

Model-led
existing coherent model -> preserve whole artefact -> document mixed basis -> selective structuring -> expose gaps -> strengthen progressively

The next operational test should use the repository owner's existing public-information assumptive journey analysis. The model should enter intact, with its public, practitioner, system-derived and analyst-synthesised basis made visible. Individual objects should be extracted or linked selectively rather than requiring the whole model to be rebuilt from atomic evidence.

Direct lived-experience research remains an important strengthening and challenge route, but it is not an entry gate.

AI continuity safeguard

The correction also records that repository-aware agents must inspect the existing non-linear ingestion, maturity and safe-ingestion guidance before recommending an entry route.

Agents should not:

• treat the latest ingestion method as the only method;
• silently narrow an accepted operating model;
• require the user to restate already documented principles;
• atomise coherent models before preserving their system-level value;
• transfer repository-governance continuity back to the user.

This requirement is also recorded in public issue #86 and private orchestration issue #10.

Object decisions

The review recommends:

• keep the five evidence records;
• keep PP_002 at draft, evidence_linked, weak, low, needs_review;
• do not create or link a user need from the Ofsted source yet;
• do not link the current evidence set to BEH_009;
• do not create a civic need until normative-basis work advances;
• treat the waiting-harm insight as an analyst inference requiring further testing;
• do not change schema or relationship types yet.

Additional findings

The review identifies, but does not fix:

• EVID_001 and EVID_002 are broader than their current waiting-stage metadata suggests;
• EVID_002 actor scope should distinguish the exact extract from adjacent family context;
• all five evidence records currently appear to support PP_002 equally, although their claim types differ;
• PP_002 and BEH_009 still link to deprecated UN_010;
• the broad object's severity is not established by the current evidence slice.

What did not change

• No evidence, need, behaviour, pain point, civic need or insight object was changed.
• No status, confidence, evidence strength, severity or review state was changed.
• No new research object was created.
• No schema or ontology relationship was added.
• No protected orchestration method was published.
• The partial-fit findings about the Ofsted evidence remain unchanged.

Recommended next step

After review and merge, run the first model-led ingestion test using the existing public-information assumptive journey analysis. Preserve the whole model, classify its mixed basis, reconcile it against current CDI objects and identify where stronger evidence would have the greatest value.

Related: #87, #86, #79, PR #85, happygeneralist/Happygeneralist#11, happygeneralist/design-intelligence-orchestration#10.