E1 ยท Formalizing: audit
Audit what the run supports
E1 checks the completed analysis against its records. It separates what can be recomputed from what the aggregator remembers or believes, then judges whether the model links population claims to local evidence well enough.
Where this stage starts
The aggregator can inspect the outputs, logs, saved versions and handoffs of the whole run, including D6. Earlier files stay as they were. An error found now becomes a recorded defect or a new correction file, and the evidence under audit stays unchanged.
What the aggregator does
Scripts recompute counts, trace quotations and identifiers, and check whether the required steps happened in the prescribed order. A separate check that the handler supplies compares quoted material with the raw record. The audit reports that check apart from the resolution of citations within the coded store.
The aggregator examines how the model developed: which cases changed successive stories, what readers were given, which category revisions occurred, and which later claims use each stage's outputs. A file that was produced and never used may point to unnecessary work or a broken handoff.
It then opens the predictions sealed before the return to communities and compares them with the answers. The comparison stays within the sample, because those communities helped build the model. First global fit, current local adequacy and support for shared forms are reported separately, with versions and unresolved evidence.
For a seeded sample of thirty communities, fresh readers judge whether the population model plus the supported local adaptations can reconstruct the claims in the community account. A poor reconstruction can come from missing evidence, an omitted local detail, an unsupported claim or a weak shared model. The audit looks for the cause in each case, and no single score gives the verdict.
Finally, the aggregator writes its experience of the run against the audited record and says where it cannot remember. New D6 proposals stay marked as unvalidated.
What the next stage receives
E2 receives a reproducible audit, the study's limits and the aggregator's separate account of its experience. The judgment on composition may be supported, unsupported or unresolved, with its level and scope stated. E2 can then examine the instructions and handling that shaped the work.
What the output files contain
Markdown files hold prose; JSONL files hold one structured record per line.
The audit and reproducible calculationsaudit-E1.md, audit-E1/, claims-E1.md, probes-E1.md
audit-E1.md reports checks, limits and the account of
methods. audit-E1/ holds scripts and outputs for replay.
claims-E1.md checks whether each claim received the kind of
evidence it needs: source words or a pattern across cases.
probes-E1.md reports recall checks and separates forgotten
information from questions the record could not answer.
Expectations checked against returnsexpectations-E1.md
expectations-E1.md compares the sealed predictions with
the communities' answers, including claims that came from the Field
account. It keeps an expected absence apart from something a case could
not test.
The local reconstruction testreconstruction-sample-E1.jsonl, RETURN-E1.md, reconstruction-E1.jsonl
reconstruction-sample-E1.jsonl records the thirty
selected communities before inspection. RETURN-E1.md
specifies the reader task. reconstruction-E1.jsonl keeps
the results and ties each to a particular claim.
Experience and the final handoffexperience-E1.md, notes-E1.md, handback-E1.md, run_manifest.json
experience-E1.md is the aggregator's first-person
account, kept apart from computed evidence. notes-E1.md
explains audit decisions. handback-E1.md carries problems
and lessons forward. The last run_manifest.json entry
records the experience file and the model that wrote it.
Full stage description
Read the complete technical card for exact tasks, reading boundaries, checks and closing rules. The explanation above is a guide; the card holds the instructions.
Show how the stage is divided into its 5 legs
E1a: scripts. Repeats: no. Interview point: -.E1b: audit section. Repeats: yes. Interview point: -.E1c: reconstruction build. Repeats: no. Interview point: -.E1d: reconstruction check. Repeats: no. Interview point: -.E1e: experience. Repeats: no. Interview point: -.
Source versions and hashes identify the exact technical documents reproduced here.