D5 · Forming: branch accounts
Explain what each proposed form actually does
A cluster label is not yet an account of an organizational form. D5 returns to the member communities and asks how each group organizes, what its members share and where that explanation breaks down.
Where this stage starts
D4 has fixed branch definitions and cross-links. The aggregator has the local community accounts, their adaptations, the population model and its limits. It can inspect the records behind a branch, but it cannot redraw the branches or change the model here.
What the aggregator does
For each branch, the aggregator writes a narrative a stranger to the model could follow. It explains what the group shares with the population account, what it does differently, and which participant evidence supports those differences. Representative cases include a poor fit, so a neat label does not conceal variation.
The account then asks whether the proposed form links constructs through a defensible explanation. A group whose shared codes cannot be connected to how its members organized remains a grouping. Branch size alone does not make it important, and a name that fits the codes but not the communities needs criticism.
Comparisons of absence require care. If a relationship is rare throughout the collection, failing to observe it in a small branch may tell us little. The stage records whether there was enough opportunity to test the claim before treating an absence as a boundary condition.
It also keeps three judgments apart: whether the original population model fit, whether the revised local account is supported, and whether the shared branch explanation is supported. A failed original fit can coexist with an adequate later local account.
Finally, the aggregator follows one explanation down to a community and its quotations, then back up. It examines how long coordination remained visible and what the records say survived. Silence stays unknown; canvas remains do not by themselves prove that organizing continued.
What the next stage receives
D6 receives the branch accounts and their limits, ready to be woven into the whole model. Suggested renamings or new distinctions travel forward as proposals. If no shared branch is supported, the account proceeds directly from the global model to local evidence.
What the output files contain
Open a file group for its purpose. These are descriptions of the stage’s outputs. Markdown files hold prose; JSONL files hold one structured record per line.
The account of each branchforms/<id>-account.md
forms/<id>-account.md explains members’
organizing, shared changes, representative and difficult cases, nearby
branches, and reasons to doubt the proposed form. It distinguishes
evidence for common codes from evidence for what participants actually
did.
A comparable record of branch judgmentsbranches-D5.jsonl
branches-D5.jsonl records branch properties, model
versions, local uncertainties and the shared explanation’s status. Its
structured format allows later checks without replacing the narrative
argument.
A path from the model to a quotationwaterfall-D5.md
waterfall-D5.md walks through the global account, a
branch where supported, a local account, its adaptations and source
words. It states what each level adds and whether any step contradicts
the one above.
What continued and what remainedsurvival-D5.md, findings.jsonl, defects.jsonl, notes-D5.md
survival-D5.md reports observed coordination, later
discussion and evidence about work kept or lost.
findings.jsonl records claims; defects.jsonl
records problems. notes-D5.md carries unresolved issues and
proposals into D6.
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.
How the stage is divided into legs
D5a: branch. Repeats: yes. Interview point: -.D5b: waterfall and survival. Repeats: no. Interview point: -.
Source versions and hashes identify the exact technical documents reproduced here.