Representation Audit

The purpose is to test whether a representation is used within the provenance, selection, medium, transformation, context, and documentary strength that can be traced. The audit may be applied to photographs, film, audio, text, quotations, transcripts, screenshots, maps, diagrams, graphs, illustrations, model images, reconstructions, animation, visualisation, and composite media products.

Representation Audit does not determine in advance whether a presentation is true, false, manipulated, authentic, or misleading. It distinguishes what the representation carries traces of, what is interpreted into it, what has been reconstructed, and what conclusions are drawn beyond what is visible or documented.

Representation Audit becomes imprecise when event, trace, registration, medium, image, text, and interpretation are used as though they were the same stage. The distinctions below are functional: one artefact may perform several roles, but those roles must then be identified separately.

That something is represented does not make it arbitrary or devoid of content about actuality. A representation may carry real traces and document bounded conditions. What it documents, however, depends upon how the trace was registered, preserved, transformed, presented, and connected to a claim.

The mandate must identify the particular representation, version, source, and claim to be tested. It is not sufficient to state that the audit will determine whether an image is real, whether a recording shows the truth, or whether a narrative is propaganda without bounding which properties and inferences are actually under examination.

The same representation may be authentic as a file, accurate as a selection, and still insufficient for a broad conclusion. Conversely, editing, compression, or captioning may be legitimate where the changes are known, relevant, and not presented as something else.

The representation should, as far as practicable, be traced from event or object through registration, storage, copying, transformation, editing, labelling, distribution, and presentation to the final claim. A strong chain does not automatically produce a strong interpretation, but a broken or unknown chain limits what can be documented.

Representations may be altered without making their content false, and they may be technically unaltered while the assigned interpretation is wrong. Provenance, integrity, and semantic validity must therefore be tested as distinct questions.

• Record who or what registered the trace, with which equipment, settings, and protocol.
• Document time, place, field of view, field of sound, light, scale, orientation, and relevant environmental conditions.
• Distinguish original from copy, export, transcode, screen recording, and published version.
• Register cropping, cutting, montage, stabilisation, colour adjustment, noise reduction, interpolation, and other transformations.
• Record captioning, speech-to-text, translation, labels, arrows, headlines, and editorial framing.
• Compare metadata with content and external grounds; discrepancies must neither be concealed nor overinterpreted.
• Document which stages are unavailable for examination and how this limits the finding.

The questions must be adapted to the representation type and consequence of use. They are not a checklist that automatically produces suspicion, but a structure for making grounds and inferences visible.

Every medium makes some conditions visible and others unavailable. A photograph freezes a field of view, audio records selected frequencies, a map reduces space to symbols, a diagram arranges values through selected axes, and text sequences statements. These conditions are not necessarily errors; they are part of what the representation can and cannot bear.

The audit must distinguish legitimate processing from alteration that makes the presentation misleading in its particular use. The assessment must consider purpose, disclosure to recipients, irreversible loss, and whether the change affects what is being claimed.

Representation capture is not another name for the human use of images, texts, maps, symbols, or narratives. The risk becomes relevant where a particular presentation gains governing priority over the matter it concerns, determines what may count as relevant counter-material, and protects that priority against correction.

An F6 finding may be recorded where concrete indicators point toward such governance, while decisive conditions for F7 remain undocumented. An F7 finding requires the relation among representation, substitution, governing priority, resistance to correction, and consequence to be documented within the mandate.

Representation Audit uses the common finding categories F1–F7. The finding must state which representation, version, chain, claim, and use were assessed, what was observed, what was interpreted, and what uncertainty or limitation remains.

A finding must distinguish technical integrity, documented origin, visible content, context, and interpretation. That a file is authentic does not by itself determine what it shows; that an interpretation is uncertain does not make the representation false.

Correction must correspond to the bounded finding. The purpose is not to demand representation-free communication, but to restore visible provenance, precise selection, relevant context, traceable transformation, and restrained assignment to actuality.

The response must address the stage to which the finding belongs: registration, storage, editing, labelling, presentation, interpretation, or decision use. A misleading headline is not necessarily corrected by changing the original file, and a broken source chain is not repaired by a footnote alone.

• Make the original, version history, and source chain available where this is responsible and lawful.
• Separate registered material from reconstruction, illustration, simulation, and artistic interpolation.
• Disclose crop, cutting, compression, colour alteration, captioning, and other relevant interventions.
• Add missing time, place, scale, orientation, axis, unit, and comparison grounds.
• Distinguish clearly among what is visible, what is interpreted, and what is claimed about cause or whole.
• Publish correction or replacement with a visible trace to the earlier presentation where the consequence requires it.
• Allow affected persons and relevant sources to present original material, context, or documented counter-material.
• Set responsibility, deadline, rollback, review, and closure criteria.

Representation Audit may itself become capturing if the auditor’s preferred crop, alternative narrative, or technical suspicion is made incorrigible. The audit must therefore keep its own selection, language, presentation, and interpretation open to the same examination.

The method may disclose provenance, selection, transformation, and inferential leaps, but cannot by itself decide every question of forgery, identity, intention, copyright, privacy, or evidential value. Such conclusions may require forensic, legal, journalistic, historical, or other professional competence.

A Representation Audit must be documented in the common audit record or an equivalent traceable format. The record should show mandate, representation version, source chain, medium, interventions, context, claim, alternative interpretations, findings, contradiction, correction, and planned review.

Review must examine whether correction actually made provenance, context, selection, and inference clearer, and whether the new presentation reduces risk without creating a new representation that conceals other relevant conditions.