Data · gold open access
Every load-bearing artefact of the book is deposited, versioned and openly licensed: CSV with a data dictionary, JSON with a validating schema, graphs in standard formats, identifiers compatible with Constitute and the Comparative Constitutions Project.
Status: repository in preparation. The first frozen release will carry a Zenodo concept DOI; until then nothing here should be cited as a validated dataset.
What the companion holds
Weighted by how much each one carries the chain: the two that build the bridge from ontology to data carry most.
Ideology × domain × register: five sub-questions per ideology family, double-coded, with Krippendorff’s α per sub-question.
Which constitutional variables count as registers of belonging, and when a change counts as a change of meaning.
Ideological families since the Enlightenment as a dated graph; constitutional borrowing as Anderson’s modular piracy, measured.
One ideology per domain per state; Mexico 1994–2018 as the worked case; a typology by clustering.
NLP on Constitute, survival models, agent-based simulation — derived, and only as strong as the coded units they rest on. Simulation as existence proof only.
Political-constitutional thought as a queryable graph; the project’s earlier proof of concept, extended.
The fused record
Mapping decisions, interpretive revisions as events, and the audit trail that links a claim to its data primitives — together, never as three separate write-ups. This is the material proof that computation and social ontology were built jointly.
{
"unit_id": "T-MEX-1917-027-a",
"source": {
"corpus": "constitute",
"constitute_id": "Mexico_2015",
"article": 27,
"clause": "a",
"span": [
0,
118
]
},
"ontology_commitment": {
"domain": "economic",
"register_mode": "who_controls_inscription",
"object_type": "social"
},
"coding_decision": {
"sq": "SQ3",
"code": "community_to_individual",
"coders": [
"LB",
"<coder-2>"
],
"codebook_version": "0.3.0",
"codebook_hash": "<sha256>",
"alpha": 0.81,
"grade": "conclusion"
},
"human_interpretation": {
"text": "The 1992 reform shifts control of the economic register without changing who is inscribed.",
"author": "LB",
"state": "interpretation"
},
"revision_events": [
{
"date": "<iso-date>",
"freeze": "F1",
"from": null,
"to": "formal_amendment",
"reason": "no change of subject"
},
{
"date": "<iso-date>",
"freeze": "F2",
"from": "formal_amendment",
"to": "reinscription",
"reason": "codebook 0.3.0 added 'who controls' as a criterion"
}
],
"lineage": [
"constitute/raw",
"segment(article,clause)",
"code(SQ1-SQ5)",
"alpha(pass 2)",
"promote(ccp_layer)"
],
"evidence_ids": [
"E-0412",
"E-0413",
"CB-0.3.0"
],
"normative_claims": "forbidden_here"
}
Every unit carries a source span with a stable id, an ontology commitment (domain and register mode), a coding decision with coders, codebook version and α, and a human interpretation marked as such. Normative claims are forbidden in the record.
Field names, the α bands (0.67 exploratory · 0.80 conclusion · below 0.67 stop and recode), the freeze points F1/F2 and the lineage vocabulary are the author’s design choices, published so they can be criticised.
Exploratory readings are open to anyone. A canonical reading needs two trained coders and an α pass; anything else must not be presented as one.
Licences and identifiers