Skip to content

GCRI · Foundations

Public-safe technical reporting: evidence without false authority

Technical outputs look authoritative, and in risk work a model or an exercise result mistaken for a warning or a decision can do harm. This page explains how The Global Centre for Risk and Innovation (GCRI) reports complex evidence so that it stays labelled, uncertain where it is uncertain, and open to correction.

As of
21 October 2026
Reading
10 minutes
Evidence
15 sources, cited in order
NEXUS TWIN · WF-1 · NETWORKSEED 15d13100GCRI-MAIN.ESSAY.OPEN-SCIENCE-FOR-RISKxyzSCHEMATIC · NETWORKST-01 · 0.801 0.366ST-02 · 0.213 0.334ST-03 · 0.940 0.328
WF-1 · network Schematic wireframe twin (network, decision under uncertainty). An illustration, not data.

Methods and evidence

Evidence
15 sources, cited in order and listed under References
Figures
4 figures
Data availability
Every chart's data is published as a table on this page.
Labels
9
evidence labels, one on every finding2
Trust
n = 5,780
participants in experiments on communicating uncertainty3
Likely
66 to 100%
the probability range of "likely" in IPCC usage5
Releases
0
approved Reports editions to date1

01What public-safe technical reporting means

What public-safe technical reporting means

Public-safe technical reporting is the practice of translating complex technical evidence, data, models, simulations and exercise results, into public statements that keep their labels, methods, uncertainty and limits, so that no reader mistakes a finding for an authority's decision, a warning, a rating or an endorsement. At The Global Centre for Risk and Innovation (GCRI) it governs Reports, the publication pillar of the Nexus Consortium that GCRI runs with The Global Risks Forum (GRF) and The Global Risks Alliance (GRA): one governed publication programme of quarterly country editions, an annual flagship and specialist series, each with its sources, limits and correction history 1,6.

"Public-safe" does not mean simplified until it is wrong, and it does not mean withheld. It means that whatever reaches the public is safe to act on in the way it says it can be acted on, and no further. A modelled flood depth is useful to a city planner if it says it is modelled, for which scenario, with what uncertainty; the same number without its label can do harm.

02Why technical evidence can create false authority

Why technical evidence can create false authority

Technical outputs look authoritative. A map with a colour ramp, a dashboard with a needle, a number with three significant figures: each carries an air of certainty that its method may not support. When such outputs come from a recognised institution, readers can take them as an institutional position, a forecast or a warning. In risk work the cost of that error is concrete. Warnings are issued by competent authorities through their own systems; in 2025, 119 countries, or 60% of all countries, reported a multi-hazard early warning system, and the global effort to cover everyone by the end of 2027 treats warning dissemination and communication as one of its four pillars 7,8. A research output that looked like a warning would compete with those systems instead of supporting them.

The public wants scientists to speak: in a 68-country survey, 83% agreed that scientists should communicate about science with the public, yet scientists scored lowest on openness to feedback, 3.33 on a five-point scale 9. Public-safe reporting answers both findings. It communicates more, not less, and it makes every statement open to challenge.

03The public-safe reporting chain

The public-safe reporting chain

Every public statement passes through six steps, and each leaves a record a reader can check. Source: every dataset and document has a named owner, a date and terms of use. Label: every finding carries one of nine evidence labels 2. Method: the method, uncertainty and limits are stated with the finding, as the series method requires 1. Review: series editors and independent reviewers check each edition, and no funder, sponsor or member can buy or block a finding, as the editorial governance page sets out. Publish: each edition is versioned and carries a persistent identifier 1. Challenge: any reader can ask for a correction, and an accepted correction produces a new version while the earlier one stays on record 1.

Fig. 01Six steps from evidence to a public statement

Six steps from evidence to a public statement Six connected panels: source with a named owner and terms, label with one of nine evidence labels (highlighted), method with uncertainty and limits, review by editors and independent reviewers, publish as a versioned edition with an identifier, and challenge, with a loop back for corrections. CHAINSourceLabelMethodReviewPublishChallenge
The chain runs from a named source to a versioned edition; a correction returns to the source and produces a new recorded version.1,2Sources: Nexus documentation, Reports and Status and service levels. As of 21 October 2026.
Figure 1 in words

Six steps left to right. 01 Source: named owner, date and terms of use. 02 Label, highlighted: one of nine evidence labels. 03 Method: method, uncertainty and limits stated. 04 Review: editors and independent reviewers; no one buys a finding. 05 Publish: versioned edition with a persistent identifier. 06 Challenge: anyone may ask for a correction, and a new version is recorded. A dashed arrow loops from Challenge back to Source with the note that a correction produces a new recorded version and the earlier version stays visible.

04Evidence labels and calibrated language

Evidence labels and calibrated language

The nine labels are observed, published statistics, institution-reported, estimated, modelled, forecast, scenario, expert judgement and unknown 2. The label "published statistics" names a source type, data reported by an authority; it is not an endorsement of a finding by that authority 2. The labels answer the reader's first question, how was this number obtained, before the number itself.

Labels say where a number came from; calibrated language says how sure the authors are. The Intergovernmental Panel on Climate Change (IPCC) gives the model most widely used in risk science. Its guidance on the consistent treatment of uncertainties assigns each likelihood term a probability range: "virtually certain" means 99 to 100%, "very likely" 90 to 100%, "likely" 66 to 100%, "about as likely as not" 33 to 66%, "unlikely" 0 to 33%, "very unlikely" 0 to 10% and "exceptionally unlikely" 0 to 1% 5. A separate scale of confidence, from very low to very high, summarises the evidence and agreement behind a finding and is not read as a probability 5. Reports adopts the same discipline: a term with a defined meaning is used only in that meaning.

Fig. 02What "likely" means when scientists use it

What "likely" means when scientists use it Horizontal range bars of the IPCC likelihood scale: virtually certain 99 to 100 per cent, very likely 90 to 100, likely 66 to 100, about as likely as not 33 to 66, unlikely 0 to 33, very unlikely 0 to 10 and exceptionally unlikely 0 to 1. Virtually certain100 per cent probability (upper bound)Very likely100 per cent probability (upper bound)Likely100 per cent probability (upper bound)About as likely as not66 per cent probability (upper bound)Unlikely33 per cent probability (upper bound)Very unlikely10 per cent probability (upper bound)Exceptionally unlikely1 per cent probability (upper bound)
Each bar spans the probability range of a calibrated term; confidence is a separate qualitative scale and is not shown.5Sources: IPCC Guidance Note on Consistent Treatment of Uncertainties (2010). As of 21 October 2026.
Figure 2 in words, and its data

A horizontal axis from 0 to 100 per cent probability. Bars: virtually certain, 99 to 100 per cent; very likely, 90 to 100; likely, 66 to 100; about as likely as not, 33 to 66; unlikely, 0 to 33; very unlikely, 0 to 10; exceptionally unlikely, 0 to 1. Source: IPCC guidance note on the consistent treatment of uncertainties, 2010, Table 1.

TermProbability of the outcome
Virtually certain99 to 100% 5
Very likely90 to 100% 5
Likely66 to 100% 5
About as likely as not33 to 66% 5
Unlikely0 to 33% 5
Very unlikely0 to 10% 5
Exceptionally unlikely0 to 1% 5
Data of Figure 2. The IPCC likelihood scale 5.

Fig. 03Nine labels, one on every finding

Nine labels, one on every finding Grid of nine evidence labels: observed, published statistics and institution-reported grouped as reported or measured; estimated, modelled, forecast and scenario as derived, with scenario highlighted; expert judgement; and unknown, shown as missing and never as zero. GroupReadingStrong linkPartial link
The nine evidence labels as recorded on 21 October 2026, grouped by how the evidence was obtained; descriptions paraphrase the definitions.2,10Sources: GCRI, Programme status; Nexus documentation, Status and service levels. As of 21 October 2026.
Figure 3 in words

Nine cards in three rows. Reported or measured: Observed, measured or seen directly with the instrument or observer named; Published statistics, data reported by an authority, a source type and not that authority's endorsement of a finding; Institution-reported, stated by an institution about itself or its systems. Derived: Estimated, from data by a stated method with its uncertainty; Modelled, produced by a model and valid within it; Forecast, a statement about the future from a forecasting method; Scenario, highlighted, a conditional what-if result, never a prediction or an observed impact. Judgement: Expert judgement, the reasoned view of named experts where data are thin. Absent: Unknown, shown as missing and never as zero.

05Stating uncertainty without losing trust

Stating uncertainty without losing trust

A common fear is that admitting uncertainty erodes trust. The evidence does not support it. In five experiments with a total of 5,780 participants, including a preregistered national replication and a field experiment on the BBC News website, people did perceive more uncertainty when it was communicated, but trust in the numbers and in the source fell only a little, and mostly when the uncertainty was expressed in words rather than numbers 3.

The same research group's review of how uncertainty is communicated separates direct uncertainty, about the quantity itself, such as a confidence interval, from indirect uncertainty, about the quality of the evidence behind it, and finds that verbal expressions such as "likely" are read inconsistently across people while numbers reduce that variability 11. Public-safe reports therefore give ranges as numbers where they exist, state the quality of the evidence separately, and keep verbal terms tied to their calibrated meanings.

06Reporting simulations, exercises and dashboards

Reporting simulations, exercises and dashboards

Simulations and exercises are where false authority is most likely. The programme's rules are explicit: simulated results are never reported as observed impact; public-safety exercises carry clear simulation marking and stop-work controls; and simulation messages are never presented as warnings from an authority 2. Scenario findings are published labelled as modelled, with their uncertainty, in series such as the scenario and stress-test series 1.

Alerting standards show how to mark a message so that it cannot be mistaken. In the Common Alerting Protocol, carried internationally by ITU-T Recommendation X.1303, every message has a required status: Actual, for recipients who should act; Exercise, for designated exercise participants only, with the exercise named; System; Test, which all recipients should ignore; and Draft 4,12. The programme uses no Actual value; its exercise outputs keep an exercise or scenario status wherever they travel.

Dashboards carry the same duties in a smaller space. The teaching values on the model pages are labelled "teaching values for a typical city-region, not measurements for any place", and published benchmarks are cited on the same line with their scope 13. Live third-party feeds keep their own attribution and are linked rather than copied as figures 13. Composition tools such as NXS-STU (composition, dashboards and workflows) are documented as designs, and the warning chain belongs to NXS-EWS (multi-hazard monitoring and warning) and the authorities it serves.

Fig. 04A warning, an exercise and a test are different messages

A warning, an exercise and a test are different messages Five cards for the CAP status values: Actual, for targeted recipients to act on; Exercise, highlighted, for designated exercise participants only; System, for the alert network; Test, which all recipients should ignore; and Draft, which cannot be acted on as it stands.
CAP status values as defined in the OASIS Standard; the programme never issues an Actual message, and its exercise outputs keep their status.4,12,2Sources: OASIS, CAP 1.2 (2010); ITU-T X.1303. As of 21 October 2026.
Figure 4 in words

Five cards showing the required status element of a Common Alerting Protocol 1.2 message. Actual: recipients who are targeted should act on it; marked as never used by the programme. Exercise, highlighted: only designated exercise participants should act on it, with the exercise named in the note element. System: supports the internal operation of the alert network. Test: technical testing only; all recipients should ignore it. Draft: a preliminary template or draft that cannot be acted on as it stands. A note states that Nexus Universe exercise outputs are marked as exercise or scenario wherever they travel and that the programme issues no Actual warning.

07AI-assisted, finance-facing and community-facing reporting

AI-assisted, finance-facing and community-facing reporting

AI assistance. Where AI tools help draft, translate or summarise, that use is disclosed under the publisher's AI-use statement, and every factual claim is checked against its source before publication; a model's output is never itself a source.

Finance-facing reporting. The finance and insurance chapter of the annual flagship and the finance and insurance readiness series make evidence readable to finance, but readiness evidence is not a financing or underwriting decision and no report is a rating 1,2. GRA's evidence and data standards set out the same boundary on its side.

Community-facing reporting. Local evidence enters a profile only within the scope its country agreed, protected knowledge never enters public examples or model training, and no membership is needed to raise a harm or correct public information 1,2. The responsibilities are set out in data and privacy responsibilities.

08Identifiers, versions and corrections

Identifiers, versions and corrections

An edition is a record before it is a document. Every edition carries a persistent identifier from DataCite together with a UNI under the proposed standard NS-1, contributor roles under the CRediT taxonomy, data published under the FAIR principles, and the publication-ethics practice of the Committee on Publication Ethics 1,14,15. Each edition states its evidence cutoff, and observed, modelled and estimated evidence stay apart 1.

Corrections are part of the record. A correction produces a new version of an edition, and the earlier version stays on record with the correction history, so a reader can always see what changed and why 1. The route is open to anyone through corrections and supersession at the Reports platform and corrections and concerns at GCRI. At the time of writing there are no approved releases: the reports gateway states "Current approved releases: None" 1.

09Standards and instruments for public-safe reporting

Standards and instruments for public-safe reporting

The reporting rules draw on established practice in science, publishing and alerting, and on proposed Nexus Standards for figures and communication.

InstrumentWhat it governsStatus
NRS-I2Risk communication and public-safe dashboardsProposed Nexus Risk Standard
NRS-I4Research evidence, reproducibility and open scienceProposed Nexus Risk Standard
NS-5Figures, maps and visualizationProposed foundation standard
IPCC uncertainty guidanceCalibrated likelihood and confidence languageIPCC guidance, 2010 5
Common Alerting Protocol 1.2Status marking of alerts, exercises and testsOASIS Standard, 2010 4
CRediT, ANSI/NISO Z39.104-2022Contributor roles on each editionANSI/NISO standard 15
Selected standards and instruments for public-safe technical reporting, with status as of 21 October 2026.

10What is changing

What is changing

  • Warning coverage is expanding. As more countries operate multi-hazard warning systems, research outputs must be easier to tell apart from warnings, not harder 7.
  • Numbers outperform words. Evidence that verbal uncertainty is read inconsistently argues for ranges and calibrated terms in every public format 11.
  • Openness is now a trust variable. The public rates scientists lower on openness than on competence, which makes visible corrections a public good 9.
  • Machine-readable status travels further. Status fields of the kind CAP defines are the model for labelling outputs that will be reused by other systems 4.

11One reporting discipline, three institutions

One reporting discipline, three institutions

GCRI publishes Reports and keeps the methods and publication process. GRF authorises publication of the year's results at gate G6 and explains what an exercise result can and cannot tell you, with its own corrections route. GRA publishes sector analysis and the finance and insurance readiness series through its publications 1. Each institution's mandate is on its home page: GCRI, GRF and GRA.

What this page does not claim: Reports has no approved release yet, and no edition is a government risk register, a sovereign rating or a ranking of countries. No programme output is a warning; only the competent public authority may warn, alert or command. GCRI does not regulate or certify, GRF does not regulate, finance or procure, and GRA does not lend, underwrite, rate or invest. Citing a standard or study is not a claim of partnership or endorsement.

Questions

Questions and answers

Is a Reports country edition a government risk assessment?

No. An edition is published after review within the scope the country agreed, and it is never a government risk register or a sovereign rating. Countries are never ranked.

Why do findings carry labels such as modelled or scenario?

So that a reader knows how a number was obtained before using it. A scenario result answers a what-if question and is never a prediction or an observed impact; missing data is shown as missing, never as zero.

Does stating uncertainty make reports less trusted?

Experimental evidence says the effect is small. In five experiments with 5,780 people, communicating uncertainty only slightly reduced trust, and mostly when it was given in words rather than numbers.

Can an exercise result be shared as a warning?

No. Simulation messages are never presented as warnings from an authority, and exercise outputs keep their exercise or scenario status wherever they travel. Warnings are issued only by competent authorities.

How are mistakes in a published edition handled?

Anyone may request a correction. An accepted correction produces a new version of the edition, and the earlier version stays on record with the correction history and its reason.

Terms

Glossary

Public-safe
Safe to act on in the way the statement says it can be acted on, and no further.
Evidence label
One of nine labels on every finding, from observed to unknown.
Calibrated language
Words such as likely that stand for stated probability ranges.
Direct and indirect uncertainty
Uncertainty about a quantity itself, and about the quality of the evidence behind it.
Exercise status
A marking that restricts a message to designated exercise participants, as in the Common Alerting Protocol.
Evidence cutoff
The date after which new evidence is not included in an edition.

Reuse and cite

Text CC BY 4.0. Cite the page with its date; data tables are linked under each figure.

Sources

References

15 sources, numbered in the order they are first cited
  1. Nexus documentation (docs.therisk.global) (2026-10-21). Reports. docs.therisk.global. Accessed 2026-10-21.
  2. Nexus documentation (docs.therisk.global) (2026-10-21). Status and service levels. docs.therisk.global. Accessed 2026-10-21.
  3. van der Bles, van der Linden, Freeman and Spiegelhalter, PNAS 117 (2020). The effects of communicating uncertainty on public trust in facts and numbers. www.repository.cam.ac.uk. Accessed 2026-10-21.
  4. OASIS Open (2010-07-01). Common Alerting Protocol Version 1.2 (OASIS Standard). docs.oasis-open.org. Accessed 2026-10-21.
  5. Intergovernmental Panel on Climate Change (Mastrandrea et al.) (2010-07). Guidance Note for Lead Authors of the IPCC Fifth Assessment Report on Consistent Treatment of Uncertainties. www.ipcc.ch. Accessed 2026-10-21.
  6. The Global Centre for Risk and Innovation (2026-10). Reports programme. therisk.global. Accessed 2026-10-21.
  7. World Meteorological Organization and UN Office for Disaster Risk Reduction (2025-11-12). Early warning systems reach new heights, but critical gaps jeopardize global progress. wmo.int. Accessed 2026-10-21.
  8. World Meteorological Organization (2026). Early Warnings for All. wmo.int. Accessed 2026-10-21.
  9. Cologna et al., Nature Human Behaviour 9, 713 to 730 (2025-01-20). Trust in scientists and their role in society across 68 countries. www.nature.com. Accessed 2026-10-21.
  10. The Global Centre for Risk and Innovation (2026-10). Programme status. therisk.global. Accessed 2026-10-21.
  11. van der Bles et al., Royal Society Open Science 6, 181870 (2019-05). Communicating uncertainty about facts, numbers and science. pmc.ncbi.nlm.nih.gov. Accessed 2026-10-21.
  12. International Telecommunication Union (2007-09). Recommendation ITU-T X.1303: Common alerting protocol (CAP 1.1). www.itu.int. Accessed 2026-10-21.
  13. Nexus documentation (docs.therisk.global) (2026-10-21). Method and limits. docs.therisk.global. Accessed 2026-10-21.
  14. Wilkinson et al., Scientific Data 3, 160018 (2016-03). The FAIR Guiding Principles for scientific data management and stewardship. www.nature.com. Accessed 2026-10-21.
  15. National Information Standards Organization (2022). CRediT: Contributor Roles Taxonomy (ANSI/NISO Z39.104-2022). credit.niso.org. Accessed 2026-10-21.

Record

Editorial record

Facts and programme states are given as of 21 October 2026. This page replaces an earlier version on public-safe technical reporting; it keeps that version's argument and grounds it in the evidence labels, edition records and correction rules now on record, with published research on uncertainty communication. It is reviewed at least every 180 days and whenever a release, a reporting rule or a cited figure changes.

Published by the GCRI editorial desk under the scientific and editorial integrity policy. To report an error or request a correction, use corrections and concerns; corrections are dated and listed with the page. Programme states are kept in the claims and readiness register.

Cite this article

GCRI editorial desk (2026, June 8). Public-safe technical reporting without false authority. The Global Centre for Risk and Innovation. https://therisk.global/public-safe-technical-reporting-translating-complex-technical-evidence-without-creating-false-authority/

Republish

Republish or translate this article free of charge, in print or online, under CC BY 4.0: keep the byline and The Global Centre for Risk and Innovation, link to the original and note any change. Figures, maps and twins made by the institutions carry the same licence; third-party data keeps its own credit line.

Attribution to paste
<p>Public-safe technical reporting: evidence without false authority — GCRI editorial desk, The Global Centre for Risk and Innovation. First published 8 June 2026 at <a href="https://therisk.global/public-safe-technical-reporting-translating-complex-technical-evidence-without-creating-false-authority/">https://therisk.global/public-safe-technical-reporting-translating-complex-technical-evidence-without-creating-false-authority/</a>. Republished under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).</p>

Topics

Related across the Consortium

Share

EmailLinkedInBlueskyMastodon

The Global Centre for Risk and Innovation (GCRI Canada) is a Canadian not-for-profit organisation, not a registered charity, with its registered office in Ontario.