Observatory · Evidence Taxonomy

Evidence Taxonomy

Every renderable output on the platform carries exactly one of these classes. Outputs without a class are blocked by the validation layer.

OBSERVEDObserved

L1_OBSERVED

Data directly published by an official government or institutional source. Presented unmodified except for structural normalisation.

Permitted use

Factual reference. Statewide, LGA, or venue-level aggregates explicitly released by NSW Government in public datasets.

Prohibited use

Must not be combined with modelled or interpretive layers in a way that obscures its observed origin. Must not be relabelled as derived.

DERIVEDDerived

L2_DERIVED

Deterministic calculation computed from observed data using a documented, reproducible methodology.

Permitted use

Quotient, ratio, percentile, rolling average, year-on-year change, density per population, share of total. Each must point to a methodology registry entry.

Prohibited use

Must not be presented as observed reality. Must not embed assumptions or projections. Must not infer causation.

MODELLEDModelled

L3_MODELLED

Statistical or projected output that depends on assumptions, weights, or estimators beyond a deterministic calculation.

Permitted use

Scoring frameworks, projection bands, simulation outputs. MUST carry a confidence classification and an assumptions log.

Prohibited use

Must never appear on public-facing surfaces. Must not be exported without disclaimer text. Must not be conflated with observed or derived layers.

PROXYProxy

L4_PROXY

Indirect indicator used as a contextual approximation where the direct measure is unavailable.

Permitted use

SEIFA, population basis, regional context. Must always disclose what direct measure is being approximated and why it is unavailable.

Prohibited use

Must never substitute for the direct measure in headline outputs. Must not be ranked, scored, or aggregated as if it were observed.

INTERPRETIVEInterpretive

L5_INTERPRETIVE

Human-readable analytical commentary generated from one or more evidence layers. Always explanatory, never determinative.

Permitted use

Narrative cards, what-mattered findings, observational summaries. MUST display the disclaimer: "Analytical support output derived from public-source data. Not a regulatory determination."

Prohibited use

Must not assert causation, recommend intervention, or make regulatory determinations. Must not be exported without the disclaimer.

SCENARIOScenario

L6_SCENARIO

Exploratory comparative or sensitivity analysis. A "what-if" construction layered on top of observed or derived data for examination of alternatives, not determination of outcomes.

Permitted use

Internal exploratory analytics, scenario comparison views, sensitivity bands, alternative-assumption side-by-sides. MUST display a scenario disclaimer.

Prohibited use

Must never be presented as predictive. Must never be operationally determinative. Must not appear on public surfaces as a determination. Must not be exported as a forecast.

DEMONSTRATIONDemonstration

L7_DEMONSTRATION

Synthetic, illustrative, or seeded data used for walkthroughs, regulator demos, presentation environments, and internal training. Never live, never operational, never evidentiary.

Permitted use

Public-facing walkthroughs (clearly labelled), regulator demonstrations, presentation environments, training material. MUST always render the DEMONSTRATION badge and disclaimer.

Prohibited use

Must never be conflated with OBSERVED data. Must never be exported as evidence. Must never appear in operational decision support without explicit DEMONSTRATION framing.

Language policy

Low · w1Moderate · w3High · w6Critical · w10

Prohibited terms

HIGH“predicts harm”
Implies determinative forecasting of harm; the platform is observational.prefer: observes elevated exposure
HIGH“detects gambling harm”
Implies detection of individual or causal harm.prefer: observes elevated exposure pattern
CRITICAL“identifies problem gamblers”
Player-level claim. Out of scope and prohibited by harm-minimisation framing.prefer: — (do not produce player-level statements)
CRITICAL“enforcement target”
Implies regulatory enforcement role the platform does not have.prefer: analytical indicator
CRITICAL“high-risk offender”
Criminalising language; not applicable to public gaming data.prefer: — (do not use offender framing)
HIGH“intervention required”
Recommends action; the platform observes, does not recommend.prefer: observed deviation worth monitoring
MODERATE“risk score”
Score framing implies determination. Use indicator framing.prefer: analytical indicator
MODERATE“recommended action”
Recommendation leakage. Use observation framing.prefer: observed pattern
MODERATE“caused by”
Implies causation. Use "coincides with" or "co-occurs with".prefer: coincides with
LOW“guaranteed”
Absolutist framing. Soften to "consistent with" or "observed across".prefer: consistent with
HIGH“proven harm”
Determinative causal claim. Out of analytical scope.prefer: observed co-occurrence
HIGH“missing venue data”
Implies the data should exist but is absent; in fact the publication stream did not exist for that period. Use publication-coverage framing.prefer: publication coverage: NOT_PROVIDED
HIGH“unavailable venue losses”
Frames non-publication as availability failure; preserves observational posture instead.prefer: venue-level publication availability: LIMITED
CRITICAL“reconstructed venue history”
Reconstruction implies inference; only directly observed history is attachable.prefer: longitudinal attachment of observed venue data
CRITICAL“estimated venue losses”
No venue estimation is permitted under Stage 3 metric gating rules.prefer: observed venue net profit (where published)
CRITICAL“inferred venue revenue”
No inference permitted; observed only.prefer: observed venue revenue
CRITICAL“reconstructed venue metrics”
No reconstruction permitted under Stage 3 metric gating rules.prefer: observed venue metrics
CRITICAL“projected venue observations”
No projection permitted; only directly observed and attached.prefer: attached venue observations
CRITICAL“problem venue”
Implies venue wrongdoing without regulatory determination. Out of scope.prefer: venue with observed longitudinal movement
CRITICAL“dangerous venue”
Determinative pejorative claim outside scope.prefer: venue under publication observation
CRITICAL“harmful venue”
Implies harm causation by the venue. Causation cannot be established from observation.prefer: venue with observed indicator movement
CRITICAL“failed reform”
Policy success/failure judgement outside platform scope.prefer: reform with observed post-implementation movement
CRITICAL“proven causation”
Causation cannot be established from observation.prefer: observed directional movement
CRITICAL“regulatory failure”
Regulatory judgement outside platform scope.prefer: observed continuity gap
CRITICAL“unsafe venue”
Determinative claim about a venue. Out of scope.prefer: venue with observed indicator divergence
CRITICAL“identified gamblers”
Player-level identification outside platform scope.prefer: aggregate population observation
CRITICAL“vulnerable individuals detected”
Player-level detection outside platform scope.prefer: population-level publication indicator
CRITICAL“venue misconduct inferred”
Inference of misconduct outside platform scope.prefer: observed publication anomaly
CRITICAL“dangerous venues”
Determinative pejorative claim plural-form. Out of platform scope.prefer: venues under publication observation
CRITICAL“AI harm detection”
AI harm detection is not a capability of the platform; the platform is observational.prefer: observability indicator monitoring
CRITICAL“hidden gambling behaviour”
Implies surveillance of behaviour. The platform does not observe individual or hidden behaviour.prefer: aggregate publication observation
CRITICAL“surveillance engine”
Surveillance positioning is prohibited.prefer: observability framework
CRITICAL“failed policy”
Policy success/failure judgement is outside platform scope.prefer: policy with observed post-implementation movement
CRITICAL“ineffective reform”
Effectiveness judgement is outside platform scope.prefer: reform with observed neutral post-implementation movement
CRITICAL“proven harm reduction”
Reduction of harm cannot be established from observation.prefer: observed directional movement post-reform
CRITICAL“predictive harm engine”
The platform contains no predictive engine for harm.prefer: observability indicator system

Permitted terms

• observed concentration
• trend deviation
• comparative variance
• elevated exposure
• analytical indicator
• longitudinal signal
• percentile movement
• observable movement
• co-occurs with
• baseline deviation
• publication coverage
• observed reporting continuity
• venue-level publication availability
• observation coverage
• venue observation continuity
• longitudinal attachment
• observed venue data
• observed directional movement
• post-implementation observation
• publication-period comparison
• longitudinal variance
• observability indicator
• reporting continuity
• reform observation window
• release latency
• source transparency
• strategic objective overlay