How the score works.

Every read carries a confidence signal.

A score is only as good as the evidence behind it. Every KILO assessment carries a plain confidence signal — Low, Medium, or High — reflecting how complete and how corroborated the underlying evidence is. A high score resting on thin or sparsely-corroborated evidence reads differently from one that several independent signals agree on, and the confidence label lets you tell them apart at a glance.

KILO also flags when a read is borderline. Where a parcel sits close to the line between two bands, the assessment says so plainly — labelling the result as sensitive to that boundary rather than presenting a borderline read as a settled one. When the band holds firmly, the recommendation is reported as robust.

What we don't do matters more.

KILO models cultural-resource sensitivity from publicly-published empirical correlates — the factors SHPD and licensed CRM firms use in pre-fieldwork HRS §6E-42 screening. Every input is a public agency dataset or a published academic source — cited and inspectable on the assessment itself. The categories of data behind a read are summarized on the home page.

The evidence base

That evidence base is not improvised. The cultural-resource layer is anchored to the standard published archaeological surveys of Oʻahu and to the State Historic Preservation Division’s own public release of the Historic Register — the properties records-holders publish, with the HRS §6E-43.5(f) segregation already performed. The spatial, regulatory, and land-tenure layers come from named state and county datasets. Every figure on an assessment traces back to a public, inspectable source; which layers KILO reads, and how it combines them, are its own.

A note on the built environment

One coverage note, in the spirit of showing the edges: historic-property review reaches the built environment too, and a structure crosses the National Register’s rolling 50-year eligibility horizon over time — so Hawaiʻi’s mid-century-modern architecture (the 1939–1979 “recent past” catalogued in the Hawaii Modernism Context Study, Fung Associates for the Historic Hawaiʻi Foundation, 2011) is an increasingly live §6E review category. KILO reads the SHPD and National/Hawaiʻi Register layers, which list designated properties; it does not yet signal an undesignated recent-past structure. If a site carries a notable mid-century building, that’s a review exposure worth checking directly — not one a parcel read will surface for you yet.

Why these signals hold up

The signals rest on the published archaeology of Hawaiʻi, not on intuition — the standard settlement-pattern and landscape-archaeology literature for the islands, the same body of work a licensed archaeologist’s predicted-findings analysis draws on. The citations don’t make the score authoritative — a screening read still answers to a licensed archaeologist’s fieldwork — but they place the reasoning inside the field’s own literature rather than outside it.

The specific thresholds and weights — how much each signal adds, how close a feature must be to count — are KILO’s own choices: heuristic priors, informed by that literature but not read off any single published figure, and not a peer-reviewed model. We state their epistemic status — how they’re tested today, and where they’re going — plainly below.

What we never touch

KILO never ingests, aggregates, or stores data that localizes ancestral remains — from any source. We use what records-holders publish under HRS §6E-43.5(f); we do not rebuild what they redact. The output is a sensitivity tier with the reasoning shown, never a claim of presence at a specific point.

The same restraint the federal framework uses

That restraint is the standard the federal cultural-resource framework itself works under. The Department of Defense’s own Integrated Cultural Resources Management Plans — including the Hawaiʻi Army National Guard’s 2016–2021 ICRMP (prepared by SEARCH) — are built on Section 304 of the National Historic Preservation Act (54 U.S.C. §307103), which directs agencies to withhold from public disclosure the location, character, or ownership of a historic resource wherever disclosure could invade privacy, risk harm to the resource, or impede a living religious practice. The Archaeological Resources Protection Act takes the same posture for site data specifically — ARPA §9 (16 U.S.C. §470hh) bars public release of the nature and location of archaeological resources absent a no-harm finding. KILO holds the same line by construction: it reads the parcel you bring and returns a tier with its reasoning — it never publishes a sensitivity map, never answers “which parcels are sensitive,” and never assembles a list. Those same plans formalize the cultural-landscape approach (required of Army installations by Army Regulation 200-1): significance is read through the spatial relationships among resources in their natural setting — ridge, stream, shoreline, ahupuaʻa — not as isolated points. That is the lens KILO’s substrate and viewshed signals already apply, which is why a parcel can carry elevated sensitivity from where it sits in the landscape rather than from a dot on a map.

A century of plowing leaves a mark — the score should reflect that.

A century of sugar and pineapple cultivation (1850s–1990s) mechanically disturbed the near-surface archaeology across much of Hawaiʻi’s arable land — the shallow deposits most field surveys look for — and SHPD calibrates review accordingly. Pretending the engine didn’t know this would be the bigger trust problem, so the score accounts for documented prior land disturbance. The adjustment is deliberately modest: it can move a marginal case between bands, but it never single-handedly drops a high-risk parcel to low risk.

Just as important, it is suppressed wherever the sensitivity isn’t surface-driven. Plowing is shallow; iwi kūpuna context isn’t — burial-adjacent, registered-site, and documented cultural-landscape context keeps full weight regardless of how much the surface has been worked.

What the survey KILO anticipates actually involves.

The HIGH-band action above says “commission an Archaeological Inventory Survey.” Here is what that buys, because the document has a fixed anatomy under HRS §6E-42 / HAR Ch. 13-284 review and the HAR §13-276 fieldwork standards. A licensed archaeology firm runs archival research first — the SHPD library and site files, OHA’s Papakilo and Kipuka databases, Bishop Museum site records, Māhele award indexes, historic survey maps, prior CRM reports — and writes a formal predicted findings section before anyone walks the ground. Fieldwork is then scaled to that prediction: full pedestrian transects where the evidence concentrates, GPS re-location of known features where documented disturbance makes new finds unlikely, hand-excavated test units only where a surface reading stays ambiguous. Every property gets a scaled plan map, a significance assessment under the five Hawaiʻi Register criteria, and the report closes with a project-effect determination and mitigation recommendations that are negotiated with SHPD, not merely filed.

Two things about that anatomy shape how to read a KILO score. An AIS begins with the same evidence-weighing KILO performs — its predicted-findings section is built from the public record before any fieldwork — so a KILO read is best understood as the screening-level analog of an AIS’s opening chapters, available before a firm is engaged; the licensed fieldwork remains the answer, not the appendix. And survey conclusions move over time, which is why KILO treats a decades-old “surveyed, nothing found” as context rather than clearance — even a no-find AIS is not a shortcut, since it still documents features, consults in good faith, and occupies real time on the project clock.

Most of the water layer is context. A narrow piece is scoring.

KILO surfaces a stack of water-side signals on every assessment — groundwater-protection zones, aquifer and injection-control regulation, watershed and nearshore-quality monitoring, and major ocean-discharge context. The reason to say so plainly is that almost all of it is intentionally non-scoring: it’s context for the developer underwriting feasibility, not a driver of the cultural-resource sensitivity tier.

The one exception is a narrow cesspool-related pathway. Cesspool replacement involves Act 125–driven excavation, and excavation in iwi-likely substrate is a documented source of inadvertent discoveries under HRS §6E-43.6; where the conversion is likely to be more invasive than a baseline replacement, the expected excavation footprint — and inadvertent-discovery exposure — rises, and the read reflects that. It is a bounded, capped adjustment: a cesspool must be present for it to apply at all, and no single signal flips a band on its own.

Alongside the score, every assessment also carries a Water Supply Feasibility classification — a four-tier feasibility verdict (connected · extendable · well-required · problematic) synthesised for the developer. This output is strictly non-scoring — the cultural-resource tier never moves with the feasibility classification. The companion narrative bullets cover the wastewater pathway and nearshore quality the same way.

Some parcels trigger a Hawaiʻi Supreme Court framework before any agency can act.

Ka Paʻakai O Ka ʻAina v. Land Use Commission, 94 Hawaiʻi 31 (2000), requires state agencies making discretionary decisions to complete a three-prong analysis of impacts on traditional and customary Native Hawaiian rights. The agency must (1) identify the valued cultural, historical, or natural resources in the petition area and the extent to which Native Hawaiian rights are exercised there; (2) determine the extent to which the proposed action will affect those resources and rights; and (3) specify the feasible action the agency will take to reasonably protect Native Hawaiian rights if they are found to exist.

KILO surfaces a Ka Paʻakai readiness flag on every assessment whose parcel location and cultural-resource score imply the project will route through one of four pathways the framework applies to: a Conservation District Use Permit (DLNR BLNR), a Hawaiian Homes Commission Act decision on DHHL trust land, a county SMA Use Permit, or §6E historic-property review at HIGH cultural- resource sensitivity. Each trigger names the pathway, the basis (e.g., “parcel sits in the Conservation district”), and the practical implication for the developer — additional agency review timeline, cultural impact assessment requirements, community consultation budget. Stacked triggers (e.g., a Conservation district parcel that’s also in the SMA) are flagged as stacked exposure.

Like the water-permit triggers, Ka Paʻakai readiness is strictly non-scoring — the cultural-resource tier doesn’t move with the readiness flag. The framework is a procedural requirement on the AGENCY, not on the developer directly. But its application substantially affects project timeline, agency review depth, and the likelihood that conditions get attached to permit approvals. KILO surfaces it as readiness context so the developer can plan for those costs rather than discover them mid-process.

The weights are heuristic priors — and we say so.

The honest status of the score, stated plainly — because the alternative, implied precision, is the failure mode this field is right to distrust. KILO’s rule weights are heuristic priors: starting estimates of how much each signal should matter, anchored to the published settlement-pattern and landscape archaeology above, but chosen by us — not derived from a peer-reviewed model, and not yet calibrated against a body of real inadvertent-discovery outcomes. We don’t hide that, and we don’t dress it up.

No calibrated probabilities — yet

For the same reason, KILO ships no calibrated probabilities. A calibration pipeline runs internally, but we deliberately keep its probabilities off the assessment until the reference set graduates from held-out sites to real Archaeological Inventory Survey outcomes. We would rather under-claim than present a prior dressed as a posterior — which is why what you see is a three-band tier and a plain Low / Medium / High confidence label, never a percentage.

From prior to posterior

The path from prior to posterior is built and running. Every assessment is logged; a licensed archaeologist or applicant can report what a survey actually found, which — once reviewed — feeds the next calibration cycle; and that outcome corpus accrues toward the day the score is measured against ground truth rather than expert priors. If you run surveys on Oʻahu, your outcomes are the ground truth that graduates this from a heuristic into a measured instrument — an open invitation, not a closed black box.

And the deepest objection — that a number gets used to justify skipping the survey — names the one outcome KILO is built to prevent. An absence of records is never read as an absence of resources; every band, including LOW, routes toward SHPD coordination or survey, never away from it; and the score is not a substitute for an AIS or §6E-42 review. The number exists to get a qualified archaeologist engaged earlier, with budget and timeline already planned — not to replace one.

Where the score stops.

KILO is a preliminary risk-screening tool. It is not a substitute for an Archaeological Inventory Survey (AIS), SHPD review under HRS §6E-42, Burial Council consultation under HAR §13-300, or any review by a state-licensed archaeologist. For projects requiring §6E-42 review, retain a licensed archaeologist.

Discovery of historic properties, including iwi kūpuna, is possible at any Hawaiʻi parcel regardless of the tool's output. A low score is a screening result, not a guarantee — and the app says so wherever a check returns negative.

Within those bounds, see what the engine actually returns on a real Oʻahu parcel — score, reasoning, every rule that fired with its cited evidence.