Epistemic Status Labels¶
Communication convention — instantiates Intellectual-Humility Narrative Integration
Short inline tags — observed, inferred, assumed, speculative, disputed, revised — that mark each claim's evidentiary status so readers weigh it correctly.
Epistemic Status Labels are a shared vocabulary of short tags attached to individual claims — in documents, tickets, model cards, chat — that declare how the author knows what they are asserting. Its distinguishing move is that the marking lives at the granularity of the single sentence and draws from a fixed vocabulary anyone can apply inline: saying "this is only speculative" costs one bracketed word rather than a paragraph of hedging. Where a running log records how one belief moved over time, a label is a snapshot on a claim where it sits; the convention standardizes admission so weak evidence can no longer hide inside confident-sounding prose.
Example¶
In an internal analysis document, a data scientist writes: "[observed] churn rose four points in March; [inferred] the rise concentrates in the annual-plan cohort; [speculative] the trigger is the new renewal email; [disputed] whether the email or the price change is primary — see thread." A product manager reading it three weeks later knows instantly which lines to act on and which to probe, without pinging the author. When the speculative line is later confirmed by an experiment, it is re-tagged [revised: was speculative, now inferred], and the change is legible in the document's history rather than silently overwritten. The whole team has learned to read a [speculative] tag not as sloppiness but as an honest invitation to test — which is exactly what turns the label from decoration into signal.
How it works¶
- A fixed, small vocabulary with a written legend, so every tag means the same thing to every reader.
- Tags placed inline at claim granularity, attached to the sentence they qualify rather than to a whole section.
- Updates re-tag rather than overwrite — a claim that firms up moves from [speculative] to [inferred], and the move is visible.
- An audience convention that a lower-warrant tag is an invitation to challenge, not an admission of weakness, so marking uncertainty carries no penalty.
Tuning parameters¶
- Vocabulary size — a few coarse tags versus many fine-grained ones. More tags buy precision but adoption falls as the legend grows.
- Placement granularity — per-claim versus per-section. Per-claim is exact but noisy; per-section is lighter but blurs which sentence is shaky.
- Default tag — whether an unlabeled claim is read as observed or as unknown status. The default silently shapes behavior.
- Enforcement — an optional convention versus a requirement in specific artifacts (model cards, RFCs, decision docs).
- Audience scope — internal-only versus external-facing, which changes how much interpretation the legend must carry.
When it helps, and when it misleads¶
Its strength is that it makes uncertainty cheap to state and easy to read, prevents fluent prose from masking weak evidence, and scales across many authors with no training beyond a legend. It retrofits onto ordinary writing the discipline that some languages build into grammar itself, where a speaker must mark whether a statement is known by observation, inference, or hearsay.[n1]
Its failure mode is that labels become decoration — everything tagged [inferred] out of habit — or worse, are gamed to sound rigorous. The classic misuse is treating a [speculative] tag as a liability shield: floating a claim without owning it, then pointing to the tag when it is wrong. The guarding discipline is a periodic informal check that tags track real evidence, plus a norm that a low-warrant tag obliges you to say what would raise it.
How it implements the components¶
uncertainty_admission_script— the tag vocabulary is the ready-made script: admitting low confidence costs one bracketed word instead of an awkward hedge.warranted_confidence_boundary— each tag draws the line between what is observed and what is merely assumed, marking the warrant of that specific claim.audience_interpretation_frame— the shared legend tells readers how to weigh a claim, so a marked uncertainty reads as diligence rather than incompetence.
It does not keep a claim's history or mark the learning_transfer_marker linking a confidence shift to a decision over time — that is the Confidence Update Log; labels tag a claim where it sits, they do not track how it moves.
Related¶
- Instantiates: Intellectual-Humility Narrative Integration — supplies the everyday marking convention that lets people admit uncertainty at low cost.
- Sibling mechanisms: Confidence Update Log · Public Assumption Revision Memo · Humility Norm Onboarding Module · Leader "I Was Wrong" Story · Learning-from-Error Postmortem · Mistake-to-Principle Story Template · Premortem Unknowns Round · Recognition for Belief Revision · Uncertainty Retrospective Prompt
Editorial Notes¶
Form Classification¶
Form family: Interface, Display & Cue
Rationale: Short inline tags sit beside each claim and provide a user-facing perceptual cue that changes how readers weigh observed, inferred, assumed, disputed, or revised content.
Nearest alternative: Communication, Facilitation & Learning — The legend establishes shared meaning, but the deployed mechanism is the visible claim-level status cue at the point where the reader encounters the statement.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Security Studies & Intelligence Analysis
Origin pattern: Convergent development
Present-day reach: Multi-domain
Rationale: Intelligence-writing standards cohered compact, local markers that keep observed facts, assessments, assumptions, and disputed claims from blending in prose.
Related originating lineages:
- Philosophy — Epistemic logic and philosophy of science supply the distinctions those labels encode.
Review resolution: The current reviewers agree that security_intelligence is primary. For the reported differences (reported_ambiguity, origin_mode_disagreement), the evidence supports convergent, multi_domain, and philosophy; these choices preserve materially formative origins without conflating later domain reach.
Attribution caveat: The exact label vocabulary varies across analytic, scientific, and online writing communities.
Review outcome: Reconciled after independent review; medium confidence.
Notes¶
[n1] Evidentiality — a grammatical feature, obligatory in languages such as Tuyuca and Eastern Pomo, that requires a speaker to mark whether a statement is known by direct observation, inference, hearsay, or assumption. ↩