Ontology

The ontology-discovery dialogue opened on a domain given by the user as “Ohio education funding”, with the goal of tooling and a web platform that explain the history, describe the current state of affairs, and predict the impact of proposed policy decisions — grounded directly in concrete numbers.

Four scoping questions were put first, because each one changes the class list:

1. GRAIN — chosen: district x fiscal year. Every funding figure anchors to a specific education agency and a specific fiscal year; statewide totals are derived by aggregation, never authored directly. Rejected alternatives were a statewide spine with exemplar districts, and statewide aggregates only. The distributional story is the substance of Ohio’s funding history, so the grain had to preserve it.

2. TIME SPAN — chosen: 1851 to present. Reaches back to the Ohio Constitution of 1851 and its “thorough and efficient system of common schools” clause (Art. VI, Sec. 2), through the 1914 School Foundation Program, HB 920 (1976), the DeRolph line (1991-2002), the Evidence-Based Model, the Bridge formula, and the Fair School Funding Plan. Accepted cost: the pre-1975 period has little usable per-agency numeric data, so early instances will be legal and structural rather than numeric.

3. CHANNELS — chosen: all four. Traditional districts and JVSDs; community (charter) and STEM schools; vouchers and scholarships; capital and facilities funding. Nothing was excluded. Higher education is out of scope by omission — the corpus is K-12.

4. PREDICTION — chosen: both simulation and projection. Simulation re-runs the real formula against perturbed parameters and emits per-agency deltas. Projection carries enrollment, valuation, and revenue series forward. The two compose: a scenario may perturb parameters on top of projected inputs.

A second round settled three structural questions:

• DATA SEAT — the corpus holds schema, mechanism, legal history, metric definitions, and curated exemplar agencies. Bulk per-agency-year facts (600+ agencies x ~35 years) live as committed data files under crates/ and are queried by the domain computer. Corpus nodes cite them rather than restating them. This keeps the graph human-readable and keeps commits legible as knowledge events.
• PARAMETERS — first-class. Base cost per pupil, the six special-education weights, the 20-mill floor, phase-in percentage, and the state share index each become nodes carrying a value-per-fiscal-year series and a governing statute link. Without this, a counterfactual has no node to point at.
• DOCTRINE and ACTOR — both modeled as classes. Doctrine carries the normative frames (“thorough and efficient”, adequacy vs. equity, phantom revenue, residual budgeting) that legislation and litigation are judged against; Actor carries the institutions that enact, rule, administer, and publish.

The sketch was presented as 12 classes and 25 edge types. The user amended it by adding fiscal-period as a 13th class, promoting the fiscal year and biennium from a property to a traversable node so that budget cycles (HB 110 = FY2022-23, HB 96 = FY2026-27) become anchors rather than repeated string fields.

The foundational alignment options were explained against three nodes from the confirmed sketch. Under BFO, education-agency is a continuant while litigation and fiscal-period are occurrents, and parameter is a quality. Under UFO, education-agency is a Kind while “property-poor district”, “20-mill-floor district”, and “DeRolph plaintiff” are Roles it plays contingently and simultaneously, and program is a Relator whose amount, eligibility rules, and litigation history are worth querying directly.

Thirteen classes, aligned to UFO. Each .ont.yml carries a foundational_type block with ontology: ufo and a type drawn from the UFO vocabulary.

Classes:

education-agency (ufo: kind) — an IRN-bearing recipient of education funding: city/local/exempted-village district, JVSD, community school, or STEM school funding-regime (ufo: phase) — a named statewide formula era with a lifespan formula-component (ufo: kind) — a structural mechanism inside a regime parameter (ufo: quality) — a dialable value with a per-fiscal-year series revenue-stream (ufo: relator) — a channel of money reaching an agency program (ufo: relator) — a separately-appropriated initiative outside the base formula legislation (ufo: kind) — an enacted bill or constitutional provision litigation (ufo: event) — a decided case testing the system against constitutional text actor (ufo: kind) — an institution that enacts, rules, administers, or publishes doctrine (ufo: kind) — a normative principle or analytical frame metric (ufo: quality) — a defined, computable measure over agency-year data scenario (ufo: situation)— a named counterfactual and its projected deltas fiscal-period (ufo: event) — a fiscal year or biennium as a traversable anchor

Edges (confirmed):

actor -[enacts]-> legislation actor -[decides]-> litigation actor -[administers]-> program legislation -[establishes]-> funding-regime legislation -[sets]-> parameter legislation -[creates]-> program legislation -[constrains]-> revenue-stream legislation -[amends]-> legislation legislation -[appropriates-for]-> fiscal-period funding-regime -[composed-of]-> formula-component funding-regime -[supersedes]-> funding-regime funding-regime -[assessed-against]-> doctrine funding-regime -[active-during]-> fiscal-period formula-component -[governed-by]-> parameter formula-component -[distributes]-> revenue-stream parameter -[valued-in]-> fiscal-period revenue-stream -[funds]-> education-agency program -[diverts-from]-> education-agency litigation -[challenges]-> funding-regime litigation -[invokes]-> doctrine litigation -[prompts]-> legislation education-agency -[party-to]-> litigation education-agency -[peer-of]-> education-agency metric -[measures]-> education-agency metric -[operationalizes]-> doctrine metric -[computed-for]-> fiscal-period scenario -[perturbs]-> parameter scenario -[projects]-> metric scenario -[baselines-on]-> funding-regime scenario -[horizon]-> fiscal-period fiscal-period -[follows]-> fiscal-period

Corpus depth: 45 seed instances, distributed toward hub classes.

WHY THESE CLASSES.

The domain’s irreducible tension is that money reaches a school from two sources that move independently — a state formula the legislature sets, and a local property tax base the legislature constrains but does not set. Every controversy in Ohio’s funding history is a consequence of the gap between them. The class list is built to make that gap expressible: funding-regime and parameter model the state side, revenue-stream models both sides as channels, education-agency is where they land, and metric measures the result.

parameter is separated from formula-component because the user’s goal is prediction. A counterfactual is an edit to a parameter’s value; if parameters are prose fields inside a component node, a counterfactual has nothing to bind to and no history to diff against. Separation makes the simulation surface an explicit, enumerable part of the graph.

metric is separated from parameter because they face opposite directions. A parameter is an input the state chooses; a metric is an output computed from data. Fusing them would blur the line between what policy sets and what policy produces — precisely the line the corpus exists to illuminate.

program is separated from revenue-stream because a voucher is not a revenue channel to a district — it is a diversion away from one, with its own eligibility rules, appropriation line, and litigation. The diverts-from edge cannot be stated if the two are merged. The user was offered that merge and declined it.

doctrine earns a class because the corpus must answer normative questions, not only arithmetic ones. “Was this constitutional” and “was this equitable” are the questions the history actually turns on, and they require the frames to be nodes that legislation and litigation can be evaluated against.

fiscal-period was added by the user over the 12-class sketch. It is the right call: with fiscal year as a mere property, “what happened in the FY2026-27 biennium” requires scanning string fields across every class. As a node it becomes a hub that legislation, parameters, regimes, metrics, and scenarios all attach to.

CONSIDERED AND DISCARDED.

• A single policy-instrument class fusing legislation and litigation. Discarded: the two have different properties (bill number and effective date vs. citation and holding) and opposite causal directions (prompts runs from litigation to legislation).
• Merging program into revenue-stream (offered, declined — see above).
• Dropping scenario and keeping counterfactuals in tooling only (offered, declined). Keeping it as a class means a projection is a committed knowledge event with provenance, not ephemeral tool output.
• An allocation class for the (agency, fiscal-period, stream, amount) tuple. Discarded at the data-seat decision: at district x fiscal-year grain this is ~20k+ instances, which would make the corpus unreadable. The tuple lives in committed data files under crates/; revenue-stream and metric nodes define its schema and cite it.

WHY UFO.

This is an institutional domain, which is UFO’s stated best fit, and two of its constructs do real work here that BFO’s continuant/occurrent split does not.

Roles: a single school district is simultaneously a taxing authority, a formula recipient, a voucher-losing entity, a DeRolph plaintiff, and a 20-mill-floor district. These are contingent, overlapping, and relationally defined — exactly what UFO Roles are for. Under BFO they would all collapse into one continuant with a pile of boolean fields.

Relators: the interesting objects in this domain are the relationships. An EdChoice deduction is not a bare edge from state to district — it carries an amount, an eligibility rule, a statutory basis, and its own litigation history. Modeling program and revenue-stream as Relators makes them queryable in their own right rather than reducible to endpoints.

BFO was a real contender because the corpus is partly an empirical record over time, and its object/event line is clean for litigation and fiscal-period. It was set aside because the analytical weight in this domain sits on who-plays-what-role-under-which- authority, not on distinguishing things from processes.

WHY 45.

The default of 13 would give roughly one instance per class — a schema demonstration rather than a corpus. 45 allows the DeRolph-to-FSFP arc to be traversable end to end with concrete numbers attached: multiple funding regimes in sequence, the bills that established them, all four DeRolph rulings, and a spread of exemplar agencies across the property-wealth spectrum. 70 was offered and not chosen; the difference is depth of exemplar coverage, which post-genesis epistemic commits can supply.

governance: was added on 2026-08-20, at the re-vendor that brought it into existence.

The field did not exist when this record was written. Bootstrap now asks whether one elector or several will maintain the repository and records the answer here, because the answer decides whether .yidam/sangha/, ma/<elector> branches, resolution events and the constitution are live machinery or dormant. This repository has had one elector throughout — the author, with agents acting on their behalf — and every phase since genesis has run on a phase/<name> branch off main, which is exactly the single-elector shape. Recording it is a statement of what has been true, not a change of course.

.yidam/sangha/ was scaffolded at genesis and is kept rather than removed; see its README for why it is dormant. The prelude’s own advice is that a single-elector repo scaffolds the sangha when a second elector appears, so this repository has the machinery early rather than wrongly.