fake news
Enable an agent to recognize suspected fabricated news, assess claims and provenance, and choose proportionate handling while preserving uncertainty and correction history.
Research draft, second pass
A second pass drafted this model: the structure a model of this thing needs, and what is known about it in the world. The line under this one says how the second half was obtained - researched against sources, or recalled without web access, in which case nothing here was read anywhere and every claim is a lead to verify. Unreviewed either way.
recalled by Codex without web access - no source was read
Researched by: Codex
Purpose and description
Enable an agent to recognize suspected fabricated news, assess claims and provenance, and choose proportionate handling while preserving uncertainty and correction history.
In its narrower analytical sense, fake news is fabricated information presented in the form of news with the intention of deceiving audiences about its factual accuracy.
It can be Extract checkable claims and identify the exact version under assessment.; Trace attributed sources and media to assess authenticity and context.; Compare claims with evidence and record bounded, revisable conclusions.; Connect circulating variants to alterations, missing context and corrections.; Recommend contextualization, review or other authorized handling based on evidence and likely consequences.; Reassess classifications and track whether corrections reach observed copies..
Distinguishing features
The item invites acceptance as factual reporting through its wording, format, attribution or circulation context; an unsupported assertion alone does not establish news imitation.
A central factual claim must be assessed against evidence; disagreement, political viewpoint or unfavorable coverage does not establish falsity.
Fabrication, material alteration and authentic material placed in a false context are distinguished from minor reporting errors and from claims that remain unverified.
Evidence of deceptive intent is recorded separately from factual accuracy; falsity alone does not demonstrate deliberate deception.
Satire or fiction labels are evaluated in the encountered version and context because redistribution may remove the cues that originally distinguished them from reporting.
Scope
+ News-like items and the factual claims on which their central message depends
+ Signals that content presents itself as authentic reporting
+ Evidence supporting, contradicting or leaving claims unresolved
+ Content provenance, material alterations and evidence concerning deceptive intent
+ Circulation context, assessments, corrections and handling decisions
- General models of journalism, news organizations and editorial production
- Misinformation that does not present itself as news or reporting
- Disinformation campaigns as coordinated operations across multiple items
- Satire, parody and fiction clearly presented and understood as such
- Platform governance, legal liability and copyright regimes in their entirety
- Audience beliefs and population-level persuasion effects beyond evidence linked to the item
Characteristics
- News presentation
- explicit reporting claim | implied reporting claim | clearly labeled fiction or satire | ambiguous Determines whether the item belongs within this model's news-like boundary.
- Claim assessment
- unassessed | supported | contradicted | mixed | insufficient evidence | not fact-checkable Separates evidence-based conclusions from suspicion and distinguishes factual claims from opinion.
- Materiality of disputed content
- central to the message | supporting detail | incidental | unresolved Prevents a small error from automatically determining the classification of the whole item.
- Misrepresentation mechanism
- fabricated claim | fabricated attribution | impersonated outlet | altered media | false context | misleading headline | multiple | unresolved Identifies what must be checked and what a useful correction must explain.
- Evidence concerning deceptive intent
- unassessed | insufficient evidence | supported by cited evidence | contested Avoids treating an inaccurate item as proof of the creator's or distributor's intentions.
- Version and derivation
- original candidate | copy of | translation of | excerpt of | altered from | corrected by Connects assessments to the exact content encountered and traces changes that affect meaning.
- Assessment provenance
- assessed claim → evidence, assessor, method and assessment time Makes classifications reviewable and supports revision when evidence changes.
- Observed circulation
- observed posts, shares or views within a named platform and time interval; unknown where unavailable Supports response prioritization without equating engagement counts with unique exposure or belief.
- Correction status
- none observed | correction issued | retracted | copies partly updated | propagation unknown Distinguishes correcting the source from correcting the versions still circulating.
Also called
Where this came from
wikidata · CC0 1.0
Drafted structure
Bundle to layer to finding to question, as the second pass will find it: 6 bundles · 11 layers · 17 findings · 29 questions.
News identity and boundary Establishes what item is being assessed and why it may be understood as news.
The label can otherwise collapse reporting, opinion, satire and political disagreement into one category.
Reporting presentation
Examines the item's invitation to believe it reports real events.
Claim to report reality
Record reporting cues, factual assertions and fiction or satire cues in the encountered context.
- Which features invite the audience to treat this item as factual reporting? definition
- Is this reporting, opinion, satire or fiction, and have redistribution or cropping removed relevant cues? boundary
Assessment unit
Separates the underlying story from a particular headline, clip, translation or repost.
Version-specific identity
Attach the assessment to a recoverable version and its material relationship to other versions.
- Which exact text, image, audio or video version is being assessed, and when was it captured? provenance
- Does the disputed meaning originate in the story, headline, attached media, translation or repost commentary? boundary
Claims and evidence Supports claim-level assessment before assigning an item-level classification.
An item may mix accurate facts, unsupported assertions and material fabrication.
Claim decomposition
Identifies checkable assertions and their importance to the item's message.
Central checkable claims
Separate factual assertions from interpretation and identify which assertions determine the central message.
- What precise assertions about people, events, dates or quantities can be checked independently? definition
- Would correcting each disputed assertion materially change the headline or central conclusion? boundary
Evidence assessment
Records the basis, limits and timing of each factual assessment.
Support and contradiction
Record accessible evidence and distinguish contradictory evidence from an unsuccessful search for confirmation.
- Which inspected sources support or contradict each central claim, and are they independent of the item being checked? provenance
- What conclusion does the evidence justify as of the assessment date, and what remains unresolved? measurement
- Is a claim being treated as false merely because corroboration was not found? boundary
Fabrication and provenance Examines how the item was constructed and whether its apparent authority is authentic.
False attribution, manipulated media and context substitution require different checks and corrections.
Authority and origin
Traces claimed authorship, publication and cited witnesses or documents.
Attribution authenticity
Separate claimed origin from verified origin and identify circular sourcing or impersonation.
- Can the claimed outlet, author, quotation or document be traced to an authentic source? provenance
- Do apparently corroborating reports provide independent evidence or repeat the same unsupported origin? provenance
Alteration and intent
Separates observable misrepresentation from claims about why a person created or shared it.
Mechanism and intent evidence
Document fabrication, alteration or false context, while attributing intent only where specific evidence supports it.
- What comparison establishes that a claim, attribution, image or recording was fabricated, altered or placed in a false context? measurement
- What evidence supports deceptive intent for a specific creator or distributor, separately from the content's inaccuracy? provenance
- Could the observed discrepancy reflect error, parody or an unresolved attribution rather than demonstrated deception? boundary
Circulation and context Tracks how versions and surrounding cues change as an item travels.
An authentic fragment can acquire a false news meaning through new captions, dates or attribution.
Variant lineage
Connects observed copies and identifies meaningful transformations.
Meaning-changing republication
Record additions and omissions that change what a circulating version appears to claim.
- Which observed copy is the earliest recoverable version, and what limits prevent identifying the original? provenance
- Which reposts change the date, location, caption, headline or disclosure in a way that changes the apparent claim? measurement
Exposure and consequences
Records observed circulation and specific reasons an inaccurate item may require attention.
Bounded impact assessment
Separate measured circulation, plausible consequences and documented audience responses.
- What circulation counts are observable, for which versions, platforms and time windows? measurement
- What decisions could this item affect, and which consequences are documented rather than merely plausible? measurement
Classification and correction Turns an assessment into an explicit classification and a reviewable response.
The model must support useful handling without converting unresolved suspicions into permanent labels.
Classification decision
Applies stated criteria and records uncertainty, authority and response rationale.
Evidence-bounded handling
Record whether the item meets the chosen definition and what action is justified within the agent's authority.
- Which explicit definition of fake news is being applied, and does the evidence satisfy each required criterion? definition
- Given the evidence, uncertainty and consequences, should the agent add context, request review, recommend a correction or take another authorized action? action
Correction and reassessment
Preserves assessment history and follows corrections across circulating versions.
Correction reach and revision
Link corrections to affected claims and versions, and identify evidence that should trigger a revised assessment.
- Which claims were corrected or retracted, by whom, and which observed copies reflect that change? provenance
- What new evidence or challenge should trigger reassessment, and how will outdated labels be updated? action
Evidence and external alignment What the world already says about this thing, gathered so the model can be checked against it.
A model that cannot be lined up against existing standards, identifiers and practice cannot be adopted by anyone who already uses them.
Reported evidence
Findings from the breadth pass, kept separate from the structural claims.
Check these first
Recalled without web access and unsourced; every item is a lead to verify.
- The term has no universally accepted scope; this description covers deliberately fabricated news-like content, not every use of the label.
- Deceptive intent is often difficult to establish from an item alone, and a creator's intent may differ from a subsequent sharer's intent.
- The listed kinds overlap and are not a formal taxonomy; identifiers and regulatory claims are omitted because none can be supplied confidently here.
- Which of these check these first hold for the sense of fake news this model covers, and on what evidence? provenance
Kinds and varieties
Recalled without web access and unsourced; every item is a lead to verify.
- Wholly fabricated news stories
- Fabricated stories impersonating established news outlets
- Fabricated news produced for advertising revenue
- Fabricated news produced for political influence
- Which of these kinds and varieties hold for the sense of fake news this model covers, and on what evidence? provenance
Real-world use
Recalled without web access and unsourced; every item is a lead to verify.
- Attracting visits and advertising revenue through sensational fabricated stories
- Influencing political attitudes or public debate
- Damaging the reputation of individuals or organisations
- Supporting scams through fabricated news coverage
- Serving as a category in research on information credibility and circulation
- Which of these real-world use hold for the sense of fake news this model covers, and on what evidence? provenance
Typical measurements
Recalled without web access and unsourced; every item is a lead to verify.
- Proportion of assessed news items classified as fabricated - 0-100; observed values depend on sampling and classification criteria - %
- Reach of a specified fabricated story - No generally applicable typical range - unique accounts or people exposed
- Audience belief in a specified fabricated claim - 0-100; results depend on question wording, population and timing - % of respondents
- Which of these typical measurements hold for the sense of fake news this model covers, and on what evidence? provenance
Failure modes and hazards
Recalled without web access and unsourced; every item is a lead to verify.
- Readers may mistake a news-like presentation for evidence of verification.
- Repetition and redistribution can detach fabricated claims from corrections or their original context.
- Fabricated stories can damage reputations and distort public decisions.
- Health or emergency fabrications can encourage harmful actions.
- Using the label indiscriminately can discredit accurate reporting and obscure distinctions between deception, error and satire.
- Which of these failure modes and hazards hold for the sense of fake news this model covers, and on what evidence? provenance
Regional variation
Recalled without web access and unsourced; every item is a lead to verify.
- The expression and its local equivalents differ in whether they primarily denote fabricated reporting or function as political accusations.
- Language, local media institutions and distribution channels affect how fabricated stories circulate and how readily they can be checked.
- Which of these regional variation hold for the sense of fake news this model covers, and on what evidence? provenance
Neighbouring kinds and how to tell them apart
Recalled without web access and unsourced; every item is a lead to verify.
- disinformation - Disinformation includes intentionally deceptive information in many forms; fake news, in the narrow sense used here, adopts the form of news.
- misinformation - Misinformation does not require deceptive intent; someone can unknowingly redistribute a fabricated news story as misinformation.
- satirical news - Satirical news uses fictional reporting for humour or criticism rather than to secure belief in its literal truth, although audiences may misunderstand it.
- journalistic error - An inaccurate report caused by a mistake lacks the intentional fabrication required by the definition used here.
- propaganda - Propaganda seeks to shape attitudes or behaviour and may use accurate information; it need not contain fabricated news.
- clickbait - Clickbait is designed to attract clicks through its presentation; its underlying claims need not be fabricated.
- Which of these neighbouring kinds and how to tell them apart hold for the sense of fake news this model covers, and on what evidence? provenance
What the second pass must settle
- Should the registry definition reserve fake news for deliberate fabrication, or include materially false news-like content when intent remains unknown?
- Where should the boundary fall for misleading headlines, selective omission and false context when the item's literal factual statements are accurate?
- What evidence threshold should support item-level classification when only some central claims can be resolved?
- Which existing Vercy models cover news, misinformation, disinformation and media provenance, and what should this entry inherit or link to?
- How should assessments represent culturally dependent satire cues and meaning changes introduced by translation or redistribution?