Inference

Inference is a conclusion drawn from indirect or related evidence, rather than from a direct measurement of the thing in question. In the context of harm-reduction reference data, an inferred value is one reasoned from pharmacologically adjacent findings — animal studies, structurally similar compounds, or known receptor mechanisms — when no direct human measurement exists.

It addresses a genuine gap: large parts of the pharmacological landscape, especially novel compounds and less-studied substances, have no controlled human data. Inference fills those gaps explicitly, which is why such values carry a distinct label — the alternative is presenting reasoned estimates as facts, which removes the information a reader needs to calibrate their trust.

How it is done

The process begins with identifying what is not known directly. A researcher or editor notes that no measured value exists for a specific property — a dose range, a duration window, a risk profile — and searches for the closest available evidence.

That evidence might be a study of a structurally similar compound, a mechanism known from the same drug class, or animal data extrapolated with known correction factors. The reasoned conclusion is recorded alongside its basis, and the result is marked as inferred rather than measured. On a substance record, this distinction appears in the confidence level or provenance label attached to the value.

What it cannot tell you

Inference cannot confirm that the conclusion is correct. Pharmacological similarity is a predictor, not a guarantee. Two compounds with nearly identical receptor profiles can differ substantially in potency, metabolism, duration, or toxicity. An inferred value may be directionally right and numerically wrong by an amount that matters in practice.

The specific failure mode worth naming: a reader sees a value labelled inferred, registers the existence of a data point, and treats it as functionally equivalent to a measured one. Reassurance from an inferred value is weaker than it appears. The label exists to prevent exactly that slide, but it only works if the reader notices and understands what it means.

Inference also cannot account for individual variation that a direct study would capture. Animal data or analogue reasoning produces an estimate for a hypothetical average case; people with unusual metabolisms, concurrent medications, or existing health conditions may fall outside that estimate in ways the inference cannot predict.