EDUKATE SINGAPORE · GOLD STANDARD PRACTICAL CASE STUDIES · CASE 113
A vivid AI reconstruction can illuminate history without becoming a photograph from the past.
A museum displays an AI-assisted reconstruction of a fictional harbour market but labels it as an original 1912 photograph. This Gold Standard case separates authentic sources, historical inference, creative interpretation and truthful visitor information.
Every artefact, harbour, image and visitor in the scenario is invented. The museum ethics sources inform interpretation principles, not a legal finding or external certification of an actual museum.
Chapter index
- 1. The visitor sees a photograph that never existed
- 2. An engaging reconstruction and a historical record do different work
- 3. Why a date can be correct for the story but wrong for the image
- 4. The ICOM Code makes accuracy an educational responsibility
- 5. The immediate response should correct the public learning record
- 6. Trace the meaning through the production workflow
- 7. Represented communities should not disappear behind plausible details
- 8. A better label can support curiosity rather than spoil it
- 9. Test the exhibit through a real visitor’s interpretation
- 10. Practice: label the evidence, not only the image
1. The visitor sees a photograph that never existed
A fictional neighbourhood museum prepares an exhibition about a harbour market said to have operated in the early twentieth century. The curatorial team has a small, dated sketch showing part of the waterfront, several written descriptions and an uncertain list of structures. A designer creates a convincing visual reconstruction using those materials and a generative image system.
The finished scene looks photographic: people move along a quay, shop signs line the road and a particular clock tower appears near the water. To keep the label short, the exhibition production team titles it ‘Harbour Market, 1912 — photograph’. During a school visit, a pupil copies the image into a history assignment as direct evidence of what the harbour actually looked like.
The curator later finds that no historical photograph of this scene was held in the collection. The displayed image was a modern speculative reconstruction. A Gold Standard museum can use imaginative interpretation, but it must not silently turn a modern interpretation into original historical evidence.
2. An engaging reconstruction and a historical record do different work
The actual source S1 is a contemporaneous sketch whose date and scope are known under the museum’s fictional records. S2 is a set of written descriptions, with varying reliability. S3 is the curatorial note identifying what can be inferred. S4 is the design brief for a generated illustration. S5 is the finished image. S6 is the public label that calls S5 a photograph.
S1 may support the presence or approximate appearance of some visible structures; S2 may support an account of activities, subject to source criticism. Neither automatically proves the arrangement of every market stall or person shown in S5. Generative tools can add plausible visual detail even when the evidence is incomplete.
That does not make reconstruction useless. A clearly identified reconstruction can help visitors imagine spatial relationships or appreciate an historical question. Its value changes when a modern design is represented as a preserved photograph. Visitors then receive a stronger factual assertion than the collection can substantiate.
GS-R02 defines evidence boundaries, and GS-R08 demands that the displayed claim match the actual source. Museum interpretation should illuminate uncertainty rather than hide it beneath photographic realism.
3. Why a date can be correct for the story but wrong for the image
The exhibition’s subject concerns a market that the fictional sources place around 1912. The production team takes the subject date and prints it immediately after the image title. To a visitor, that placement reads as the date when the image itself was made, not merely the era being illustrated.
A more accurate label separates production and subject: ‘Artist’s digital reconstruction of the harbour market as it may have appeared around 1912; created for this exhibition in 2026.’ It can then identify which elements draw on preserved sources and which remain interpretive. The date of the event and the date of the image are both useful, but they answer different questions.
An image that includes an unverified clock tower should not be used as direct evidence that the tower existed at that location. The team may retain it only if it is an acceptable interpretive element under the museum’s research and representation principles, with sufficient disclosure or revision.
The Gold Standard does not prohibit evocative storytelling. It requires the provenance and confidence of a claim to survive layout, editing and marketing without quietly changing meaning.
4. The ICOM Code makes accuracy an educational responsibility
The International Council of Museums Code of Ethics is a recognised professional reference for museum practice. Its provision 4.2, Interpretation of Exhibitions, asks museums to ensure exhibition information is well founded and accurate, with appropriate consideration for represented groups and beliefs.
The official Code is a sector ethics reference, not proof that the fictional museum is a member or has violated a particular legal duty. It offers useful guidance about why the integrity of displayed interpretation matters to the public, especially when a museum’s authority can cause viewers to treat a claim as historically established.
The Museums Association discussion of exhibition labels also describes labels as important mediators of meaning, evidence and voices. The exact form of a label can shape what visitors understand about an object and its relationship to the world.
The Gold Standard framework asks the museum to transform that ethical aspiration into a repeatable content review: source, interpretation, label and visitor understanding must remain connected.
5. The immediate response should correct the public learning record
The museum’s accountable curator pauses the description ‘original photograph’ while confirming how S5 was produced. Staff preserve the sketch, design brief, generated image versions, review history and public text. They should not delete the reconstruction and then claim it never appeared, especially when visitors may have used it in assignments.
An appropriate dated correction can explain that the visual was a contemporary reconstruction created from selected historical references, not an archival photograph. Where the display continues, a revised label should be visible at the point of viewing. The museum may also need to correct a catalogue entry, downloadable worksheet or online post that repeated the original description.
The school group receives an accessible explanation that allows learners to revise any claim they made from the image. The museum should not shame the student for treating a confidently labelled museum image as evidence. The institution supplied the classification that encouraged the mistake.
GS-R09 calls for an immediate restriction of unsupported claims; GS-R10 calls for correction and recurrence prevention. The purpose is to repair understanding, not simply fix typography.
6. Trace the meaning through the production workflow
The curator initially called the project ‘a possible reconstruction’. The design brief shortened the term to ‘market illustration’. The gallery designer then used a template labelled ‘historic photograph’ because the visual looked photographic. The final copy editor checked spelling but did not compare the label with the provenance record.
Each step made a locally understandable choice and the result still became misleading. The issue is not solved by instructing all designers never to use generative tools. It is solved by preserving the classification of an image across production: original archival record, modern photograph, replica, reconstruction, artistic impression or other clearly defined category.
The final approver needs to know which source items support the representation, who verified the provenance and which uncertainties should reach visitors. A designer’s successful export cannot itself approve the historical claim. A curator’s early note does not help if it vanishes before the label is printed.
GS-R03 assigns decision authority, GS-R07 requires a usable handoff and GS-R12 protects meaningful classifications when content formats change.
7. Represented communities should not disappear behind plausible details
The generated scene includes distinctive clothing and shopfront features that were not present in the actual source sketch. Some may have been introduced because the design system favoured visually familiar historical patterns. The curator cannot treat those additions as neutral just because they make the display attractive.
The museum should examine whose people, places and activities are being represented, which evidence is available and whether particular interpretations should be discussed with relevant communities or subject specialists. A fictional harbour market can still teach the principle without making claims about an actual historical ethnic group.
An interpretive reconstruction may intentionally simplify or invent visual elements. The audience can enjoy it when the institution makes the distinction clear. If the museum wants to suggest a particular cultural practice, the curatorial review should identify the evidence and its limitations rather than let a generated image decide what is true.
The Gold Standard aims for both clarity and respect. It does not reduce historical responsibility to a small ‘AI’ badge while leaving significant unsupported claims in the picture or accompanying prose.
8. A better label can support curiosity rather than spoil it
The museum drafts a new panel: ‘Imagined harbour market scene, created in 2026. This digital reconstruction draws on an early twentieth-century waterfront sketch and written descriptions. The arrangement of individual stalls, clothing and several structures remains uncertain.’
The accompanying panel separates the original sketch from the reconstruction. Visitors can compare what the museum knows directly, what it infers from multiple sources and what the design team created to make the space understandable. That distinction gives a teacher an opportunity to ask how historians work with incomplete evidence.
The revised format need not be wordy or defensive. A concise title, an explicit production description and one or two meaningful source links may be sufficient for some displays. More complex or sensitive topics may require deeper explanation and community consultation.
A Gold Standard museum can be engaging precisely because it makes evidence visible. Visitors should leave able to tell what has survived from the past and what modern people have suggested about it.
9. Test the exhibit through a real visitor’s interpretation
The review team gives a reader unfamiliar with the project the final panel, image and accompanying materials. The reader is asked whether they believe the picture itself was made in 1912, what the original sources are and which details should be treated as interpretation. If the reader cannot tell, the label needs revision.
A second test examines the website, school worksheet, social post and printed catalogue. A correct gallery label is insufficient if the online download still calls the reconstruction a historical photograph. Staff verify that the image’s classification is preserved across channels and that later reuse requires a review of the source and wording.
An exercise also introduces a new artist’s reconstruction generated from the same sketch. The institution should be able to distinguish the two modern images from each other and from the archival source rather than treat them as multiple corroborating historical records.
Passing these checks supports the tested interpretations. It cannot eliminate all uncertainty from historical research; it demonstrates that the museum now gives visitors a fair route through that uncertainty.
10. Practice: label the evidence, not only the image
Exercise. A fictional museum possesses a 1912 sketch but displays a 2026 AI-assisted reconstruction captioned ‘Harbour Market, 1912 — photograph’. A school uses it as historical evidence. Write the bounded finding, immediate correction, evidence classification and public retest.
A reasoned finding is: ‘The displayed image was created in the present from historic source material and does not constitute an original photograph of the claimed year. Its label overstates provenance and may lead visitors to treat speculative visual details as direct evidence. The museum should correct the label and related publications while preserving the underlying records.’
The future test asks whether a visitor can distinguish original source, scholarly inference and creative reconstruction without privileged staff knowledge. The Gold Standard lesson is that vivid interpretation earns trust when it carries its provenance honestly into the public’s understanding.
Source context and related Gold Standard casework
- ICOM · Code of Ethics for Museums
- Museums Association · Rethinking Museum Labels
- Museums Association · Displaying Prehistory and Interpretation
- The Gold Standard of Museums · Objects, Evidence and Meaningful Visits
- The Gold Standard of Cultural Heritage · Intelligible Public Access
- The Gold Standard · Information and Record Integrity
- Case 19 · AI Assistant Invented a Policy
- Case 50 · Museum Rights for Print but Not Online
- Gold Standard Casebook
- Master Gold Standard Hub
The ICOM Code provides authoritative international professional ethics context. Its application and governing museum duties must be assessed in the relevant institution. GS-R01–GS-R12 is eduKate’s educational management framework.