Research & AI Disclaimer

Last updated: 16 August 2026.

Gausspeak is an experimental decision-intelligence platform under evaluation. This page explains how its evidence layers should and should not be interpreted.

Measurement is not interpretation

Canonical Q/M/D values and engine classifications are stored as measured model outputs. Contextual signals and AI-generated interpretations are separate derived layers and do not overwrite the canonical measurement.

Forecasts are probabilistic model outputs

Monte Carlo trajectories describe simulated distributions under stated model assumptions. They are not observations, promises or factual statements about the future. Forecast accuracy should only be reported after a cycle-specific forecast has been persisted with provenance and compared with a later canonical measurement.

Interventions are hypotheses

Proposed interventions and expected Q/M/D changes are testable hypotheses, not prescriptions. Observed change after re-measurement may differ materially from the expected change. Intervention expectation error is distinct from forecast error and should not be presented as the same metric.

Scientific status

The current evaluation release must not be described as scientifically validated unless and until appropriate empirical validation, calibration, replication and methodological review have been completed. Regular and Deep assessment modes must remain analytically distinguishable unless empirical evidence supports an equivalence claim.

Human review required

AI-generated interpretation and recommendation text can be incomplete, uncertain or wrong. Qualified human review is required before consequential action. Gausspeak should be used to structure inquiry and test assumptions, not to eliminate judgment or accountability.

Not for individual high-risk decisions

Gausspeak is intended for organizational and research-oriented analysis. It is not intended to make or determine individual employment, medical, legal, credit, insurance or other decisions that create legal or similarly significant effects for a person. Users are responsible for ensuring that their use complies with applicable law and organizational governance requirements.

Research integrity

Research use should preserve provenance, distinguish measured from inferred variables, use documented inclusion criteria, retain reproducible dataset snapshots where appropriate, and avoid post-hoc relabelling of model outputs as observations. Pseudonymisation reduces but does not necessarily eliminate re-identification risk.

Questions about research or model interpretation may be sent to info@gausspeak.com.