Provisional assumption policy

# Habitability — Provisional Assumption Policy v1

Status: RESEARCH / CANONICAL PROPOSAL  
Version: 1.0  
Doctrine: **Assume to explore. Prove to validate.**

## Purpose

Habitability must remain scientifically rigorous without allowing every unresolved external semantic or methodological question to halt research indefinitely. This policy separates permission to explore from permission to validate or publish.

An unresolved assumption may support controlled RESEARCH work when it is explicit, justified, reversible and testable. It may never be silently treated as established evidence.

## Core rule

- **RESEARCH / PROVISIONAL:** documented assumptions may be used for controlled experimentation.
- **VALIDATED:** material assumptions must be resolved by the evidence and review gates required by the MVE contract.
- **PUBLISHED:** all validation requirements remain satisfied and the publication gate, human review, Founder approval and publication licensing remain mandatory.

A HOLD therefore blocks the evidence-state transition it affects. It does not automatically block a bounded research experiment unless the uncertainty makes even exploratory computation misleading or unsafe.

## Assumption record

Every material provisional assumption used in a computation must have a stable `Assumption ID` and record:

1. statement of the assumption;
2. reason it is needed;
3. evidence supporting its plausibility;
4. evidence gap preventing validation;
5. confidence: LOW / MEDIUM / HIGH;
6. potential impact on the result;
7. exact scope where it may be used;
8. prohibited uses;
9. sensitivity or falsification test where feasible;
10. resolution condition and authoritative source sought;
11. owner and date;
12. downstream artifacts/results that depend on it.

When an assumption changes or is resolved, dependent RESEARCH results must be identifiable and reproducibly recalculated.

## Decision classes

| Evidence condition | Controlled research | VALIDATED | PUBLISHED |
|---|---|---|---|
| Required evidence complete | permitted | eligible for validation gate | eligible for publication gate |
| Explicit, plausible, testable provisional assumption | permitted with Assumption ID | blocked | blocked |
| Material uncertainty that makes exploratory output misleading | only sensitivity/diagnostic work | blocked | blocked |
| Indefensible or unbounded source/treatment | blocked | blocked | blocked |

## H05 initial application

### ASM-H05-001 — PM2.5 numeric unit

- Provisional assumption: the 2024 annual-average PM2.5 raster values are expressed in `µg/m³`.
- Basis: official EEA material for the same PM2.5 annual concentration mapping family expresses PM2.5 annual concentrations in `µg/m³`; the exact raster band does not encode its unit.
- Confidence: HIGH for exploratory use, not sufficient for validation of the exact artifact semantics.
- Permitted use: RESEARCH / PROVISIONAL calculations and sensitivity work.
- Prohibited use: promotion of H05 to VALIDATED or PUBLISHED on this assumption alone.
- Resolution: authoritative EEA/ETC confirmation tied to the exact artifact or equivalent official artifact-level semantics.

### ASM-H05-002 — `-99.0` treatment

- Provisional assumption: cells equal to `-99.0` are not valid physical PM2.5 concentrations and are excluded from exploratory aggregation.
- This is deliberately **not** a claim about what provider-side missingness/domain class `-99.0` represents.
- Basis: PM2.5 concentration cannot be negative; direct artifact inspection found `-99.0` values distinct from the technical GeoTIFF nodata value.
- Confidence: HIGH that `-99.0` must not enter concentration arithmetic; LOW regarding its exact provider-side semantic meaning.
- Permitted use: exclusion from RESEARCH / PROVISIONAL arithmetic, with excluded-cell and affected-population coverage reported.
- Prohibited use: silently recoding `-99.0`, assigning it a provider meaning, zero-filling it, or using this assumption to promote H05.
- Sensitivity: report spatial/population coverage affected and test whether conclusions are materially sensitive to exclusion.
- Resolution: authoritative EEA/ETC definition and downstream masking/exclusion rule.

## Machine-gate relationship

This policy does not weaken the MVE gate. The existing `RESEARCH → VALIDATED` requirements remain fail-closed: no blocking gates, quality PASS, reproducibility PASS, confirmed reuse status and independent approval reference.

A provisional assumption belongs inside RESEARCH evidence as an explicit uncertainty/limitation and blocking gate. Passing structural MVE validation means the package is reviewable, not scientifically validated.

## Review Gate relationship

Reviewers must distinguish:

- whether a provisional assumption is sufficiently documented to permit bounded RESEARCH;
- whether sensitivity/materiality is adequately tested; and
- whether the assumption has actually been resolved sufficiently for evidence promotion.

A MERGE-READY verdict may therefore merge reproducible RESEARCH work containing provisional assumptions. Merge never means that the assumption or resulting indicator is VALIDATED.

## Observatory relationship

The Observatory may report that research is progressing under provisional assumptions, but must keep management completion separate from evidence status. It must not display provisional evidence as VALIDATED and should expose unresolved material assumptions where they affect the critical path.

## Reversibility requirement

Provisional assumptions are acceptable only when their downstream impact is traceable. A later authoritative answer must allow Habitability to identify affected artifacts, rerun the relevant pipeline and compare the revised result without reconstructing undocumented decisions.

This reversibility requirement is part of scientific reproducibility, not optional project documentation.