AI Policy

Franz-Josef-Land.info may use artificial intelligence tools as part of its research, editorial, organizational, and publishing workflow.

AI can be useful for:

  • organizing research,
  • identifying questions for further investigation,
  • comparing terminology,
  • summarizing notes,
  • generating outlines,
  • assisting with translation,
  • improving readability,
  • drafting explanatory text,
  • and creating non-documentary illustrations.

However, AI is not treated as an authoritative historical or scientific source.

Our central rule is:

AI may assist the work, but factual claims must ultimately be supported by identifiable evidence.

Why We Have an AI Policy

Artificial intelligence tools can process large amounts of information quickly, but they can also:

  • generate incorrect facts,
  • invent citations,
  • confuse similar place names,
  • misstate dates,
  • oversimplify scientific findings,
  • present uncertainty as certainty,
  • or produce convincing information with no reliable source behind it.

These risks are particularly relevant to Franz Josef Land.

The archipelago has:

  • complex expedition history,
  • multiple transliteration systems,
  • historical place-name variants,
  • conflicting maps,
  • incomplete archival records,
  • changing glacier boundaries,
  • and scientific datasets covering different periods.

For these reasons, AI-generated output requires verification rather than automatic acceptance.

AI Is an Editorial Tool, Not an Authority

Franz-Josef-Land.info may use AI tools to assist with editorial tasks.

Possible uses include:

  • research planning,
  • article structure,
  • topic discovery,
  • preliminary summaries,
  • source comparison,
  • language editing,
  • grammar correction,
  • metadata drafting,
  • internal linking suggestions,
  • and content organization.

AI output is treated as working material.

It does not become reliable merely because it is fluent or detailed.

The underlying evidence remains more important than the generated wording.

Human Accountability

Responsibility for published material remains with the people operating and editing Franz-Josef-Land.info.

An AI system cannot:

  • accept editorial responsibility,
  • verify its own claims independently,
  • determine historical truth,
  • or replace accountability for errors.

This principle is consistent with current scientific-publishing policies. Nature Portfolio, for example, states that large language models do not satisfy authorship requirements because authorship includes accountability for accuracy and integrity.

We therefore do not list an AI system as the author of an article simply because it assisted with drafting or editing.

Verification of AI-Assisted Content

Factual information produced with AI assistance should be verified against appropriate sources before publication whenever the claim is significant.

Depending on the topic, verification may involve:

  • original expedition accounts,
  • historical maps,
  • archival collections,
  • peer-reviewed scientific papers,
  • government sources,
  • museums,
  • polar research institutions,
  • geographic datasets,
  • satellite data,
  • or authoritative reference works.

Examples of claims requiring particular care include:

  • expedition dates,
  • latitude records,
  • explorer names,
  • island identification,
  • glacier statistics,
  • wildlife numbers,
  • historical quotations,
  • geographic coordinates,
  • and current administrative rules.

AI output itself is not considered sufficient evidence.

AI-Generated Citations

AI systems can sometimes produce references that:

  • do not exist,
  • contain incorrect titles,
  • identify the wrong author,
  • use the wrong DOI,
  • or point to sources that do not support the claim.

For this reason, citations suggested by AI should be independently checked.

A source should ideally be confirmed through:

  • the publisher,
  • DOI record,
  • archive,
  • library catalog,
  • institutional website,
  • or original document.

Fabricated references are not acceptable.

AI and Historical Research

AI can be useful in historical research for tasks such as:

  • identifying alternative spellings,
  • comparing expedition names,
  • organizing chronology,
  • generating archive search terms,
  • and assisting with translation.

However, AI should not be used to invent missing historical evidence.

If no surviving source establishes what happened, the website should not create a plausible event merely to fill the gap.

The correct conclusion may be:

the available evidence is insufficient.

AI and Historical Quotations

AI-generated wording must never be presented as a genuine historical quotation.

A quotation attributed to:

  • Julius Payer,
  • Carl Weyprecht,
  • Fridtjof Nansen,
  • Hjalmar Johansen,
  • Frederick Jackson,
  • Benjamin Leigh Smith,
  • or another historical person

should be traceable to a real source.

If the exact wording cannot be verified, we prefer paraphrasing rather than placing uncertain text inside quotation marks.

AI and Historical Maps

Artificial intelligence may assist with:

  • extracting labels,
  • comparing place names,
  • interpreting damaged text,
  • or organizing historical map information.

It should not silently reconstruct missing geographic information and present the result as though it appeared on the original map.

Any reconstruction should remain distinguishable from the historical source.

AI and Translation

AI-assisted translation may be useful when researching sources in:

  • German,
  • Russian,
  • Norwegian,
  • Italian,
  • French,
  • or other languages.

However, translations can introduce errors, particularly with:

  • historical terminology,
  • geographic names,
  • nautical terminology,
  • scientific vocabulary,
  • and ambiguous nineteenth-century language.

Where exact wording matters, the original-language source should be consulted whenever practical.

AI translation should be treated as assistance rather than definitive philological interpretation.

AI and Scientific Research

AI may help summarize scientific papers or identify concepts requiring further investigation.

It should not replace reading the underlying study when a claim materially depends on that research.

For scientific claims, important context may include:

  • study area,
  • observation period,
  • sample size,
  • methodology,
  • uncertainty,
  • model assumptions,
  • and limitations.

An AI summary can easily omit these details.

Nature Portfolio’s current AI policies similarly retain human accountability and require substantive AI-assisted material to remain subject to editorial review and fact-checking.

AI and Numerical Data

Numbers require particularly careful verification.

AI-generated figures should never be assumed correct.

Examples include:

  • glacier area,
  • glacier mass loss,
  • sea-ice decline,
  • island counts,
  • wildlife populations,
  • elevations,
  • distances,
  • and historical coordinates.

Before publication, important numerical claims should be checked against the original dataset or publication whenever practical.

AI and Geographic Names

Franz Josef Land is especially vulnerable to AI errors involving place names.

A location may have:

  • German expedition-era naming,
  • Russian spelling,
  • English spelling,
  • alternative transliterations,
  • historical variants,
  • or similar names used for different places.

AI systems may incorrectly merge two locations or treat spelling variants as separate entities.

Geographic names should therefore be compared against reliable historical or modern mapping sources.

AI and Current Information

AI should not be relied upon for current information unless the relevant facts have been checked against up-to-date sources.

This applies particularly to:

  • protected-area regulations,
  • travel access,
  • permits,
  • government administration,
  • current research stations,
  • legal requirements,
  • and contemporary scientific data releases.

A model may contain outdated information even when its response sounds current.

Human Review

AI-assisted content should receive editorial review appropriate to its importance and risk.

Review may include:

  • factual checking,
  • source checking,
  • citation verification,
  • removal of unsupported claims,
  • correction of misleading wording,
  • comparison with primary evidence,
  • and rewriting where necessary.

The objective is not simply to correct grammar.

It is to ensure that the final article reflects evidence rather than model-generated confidence.

AI Does Not Replace Sources

AI systems are not normally cited as evidence for historical or scientific claims on Franz-Josef-Land.info.

Instead, the underlying sources should be cited.

For example:

Rather than citing an AI answer stating that Nansen reached a certain latitude, the article should cite the expedition record, museum source, or scholarly history documenting that event.

Rather than citing an AI-generated summary of glacier loss, the article should cite the scientific paper or dataset.

AI Does Not Replace Expert Judgment

Certain questions require expertise.

AI-assisted explanations may help make technical research accessible, but they should not be treated as replacements for:

  • glaciologists,
  • geologists,
  • Arctic ecologists,
  • historians,
  • archivists,
  • cartographers,
  • navigators,
  • or protected-area authorities.

This is especially important when information could influence operational or safety decisions.

AI-Generated Images

AI-generated visuals may occasionally be used for:

  • decorative artwork,
  • conceptual illustration,
  • website branding,
  • non-documentary diagrams,
  • or clearly illustrative scenes.

Such images should never be presented as authentic:

  • historical photographs,
  • expedition photographs,
  • archival maps,
  • satellite images,
  • archaeological evidence,
  • or scientific observations.

If a synthetic image could reasonably be mistaken for documentary evidence, its artificial nature should be disclosed.

AI-Generated Historical Scenes

A generated reconstruction of an expedition camp, ship, or Arctic landscape may be useful for illustration.

However, it should be labeled appropriately.

It should not be captioned in a way that suggests:

Photograph of the expedition camp, 1896

if no such historical photograph exists.

A more appropriate description might be:

Illustrative reconstruction of a late nineteenth-century Arctic expedition camp.

The distinction matters because visual material can easily acquire false documentary authority.

AI Images and Research Integrity

Policies across scholarly publishing remain cautious about generative imagery because of unresolved issues involving provenance, copyright, and research integrity. Nature Portfolio currently restricts generative AI imagery in its research publications and emphasizes disclosure and human oversight for permitted AI uses.

Franz-Josef-Land.info is not an academic journal, but we adopt the underlying principle that synthetic visuals should not be confused with primary evidence.

AI Image Editing

AI tools may sometimes be used for image enhancement or restoration.

Possible examples include:

  • noise reduction,
  • scratch repair,
  • upscaling,
  • or improved legibility.

Care must be taken not to add historically significant details that were absent from the original.

If AI alteration materially changes what viewers might interpret as historical evidence, the modification should be disclosed.

The original image should remain the evidentiary source.

AI and Maps

AI may assist with:

  • label placement,
  • map design,
  • summarizing geographic datasets,
  • or converting research notes into cartographic annotations.

AI should not independently determine authoritative boundaries, coordinates, or coastlines.

Maps should remain grounded in identifiable geographic data.

Our Map & Data Sources page explains the datasets and limitations relevant to geographic visualizations.

AI and Source Discovery

AI can help generate useful research queries.

For example, it may suggest searching for:

  • alternative explorer name spellings,
  • Russian transliterations,
  • older map titles,
  • expedition ship names,
  • or related scientific terminology.

Source discovery is one of AI’s most useful roles.

But discovering a possible source is not the same as validating it.

The original record must still be evaluated.

Confidential and Unpublished Material

Sensitive, confidential, private, or unpublished material should not be uploaded to third-party AI services unless doing so is authorized and consistent with applicable privacy, confidentiality, and contractual requirements.

Scientific-publishing policies similarly warn against uploading confidential manuscripts into unsecured AI tools because doing so may compromise confidentiality.

The same principle applies to:

  • private correspondence,
  • embargoed research,
  • unpublished manuscripts,
  • restricted archives,
  • and personal information.

AI-generated and AI-assisted material can raise questions involving:

  • copyright,
  • training data,
  • derivative works,
  • image provenance,
  • and reuse rights.

Where AI output incorporates or closely imitates protected third-party work, additional review may be necessary.

AI should not be used deliberately to reproduce copyrighted text, maps, photographs, illustrations, or distinctive creative works without appropriate authorization.

Plagiarism

AI assistance does not change our originality standards.

Content should not reproduce another author’s distinctive expression without appropriate quotation or attribution.

AI-generated paraphrasing should not be used to conceal plagiarism.

The underlying principles in our Sources & Citation Policy continue to apply.

Disclosure of AI Use

Not every minor AI-assisted task requires a prominent disclosure.

Examples of routine assistance may include:

  • spellchecking,
  • grammar improvement,
  • formatting,
  • headline variations,
  • or internal organization.

More substantial AI involvement may warrant disclosure when knowing about it would materially help readers evaluate the content.

Examples could include:

  • substantial AI-generated explanatory text,
  • synthetic historical illustrations,
  • AI-generated reconstructions,
  • or machine-generated analysis incorporated into a research feature.

The appropriate level of disclosure depends on how AI was used and whether that use affects interpretation.

Editorial Decisions Remain Human Decisions

AI may suggest:

  • titles,
  • article structures,
  • classifications,
  • related articles,
  • or source leads.

Final decisions about:

  • what to publish,
  • what evidence to trust,
  • how uncertainty is described,
  • and whether corrections are required

remain editorial decisions.

NIST’s AI Risk Management Framework similarly emphasizes managing AI risks through structured governance and responsible organizational use rather than treating AI output as inherently trustworthy.

Corrections of AI-Assisted Errors

If AI-assisted content introduces an error, it is treated like any other editorial error.

The website may:

  • correct the factual statement,
  • replace or remove an unsupported citation,
  • revise misleading language,
  • remove an inaccurate image,
  • or add a correction note where appropriate.

Responsibility does not shift to the AI system.

Our Corrections Policy governs substantive changes.

AI and Reader Submissions

If future website features allow users to submit AI-generated material, those submissions may be subject to additional review.

AI-generated submissions should not:

  • impersonate historical documents,
  • fabricate citations,
  • falsely claim authorship,
  • infringe third-party rights,
  • or deliberately misrepresent synthetic material as authentic evidence.

Automated Tools and Databases

Not every automated system is generative AI.

The site may also use conventional software or machine-learning tools for:

  • search,
  • categorization,
  • geographic processing,
  • image analysis,
  • OCR,
  • duplicate detection,
  • or database maintenance.

These tools are evaluated according to their purpose and risk.

Automation alone does not remove the need to verify important outputs.

Our Standard for AI-Assisted Publishing

Before publishing content substantially assisted by AI, we aim to ask:

Are the factual claims supported by identifiable sources?

Have citations been checked?

Are quotations authentic?

Are dates, numbers, and coordinates verified?

Has uncertainty been preserved?

Could any generated image be confused with historical evidence?

Has confidential material been protected?

Would disclosure of AI use materially help the reader understand the content?

Has someone taken responsibility for the final result?

If those questions cannot be answered satisfactorily, the material may require further review before publication.

Policy Review

Artificial intelligence technology is evolving rapidly.

Industry, academic, legal, and editorial standards continue to change as new systems and risks emerge.

Nature Portfolio, for example, describes its AI policies as subject to continuing review as technology develops.

Franz-Josef-Land.info may therefore revise this policy as:

  • AI capabilities change,
  • legal requirements evolve,
  • publishing standards develop,
  • or new risks become apparent.

Our Commitment

Franz Josef Land has been documented through generations of improving technology.

Early explorers used:

sextants, compasses, journals, and sketches.

Later researchers used:

radio, aircraft, photography, and scientific instruments.

Modern researchers use:

satellites, digital archives, GIS, remote sensing, and large scientific databases.

Artificial intelligence is another tool within that continuing technological progression.

But a better tool does not eliminate the need for evidence.

Franz-Josef-Land.info uses AI with the same principle applied throughout the project:

technology may help us find, organize, explain, and visualize information, but it should not replace sources, accountability, or critical judgment.

When evidence and generated output disagree, the evidence comes first.