Gains:
- Ability to use AI to produce accessible exhibition text, multilingual access and storyboards, prioritizing accuracy over appeal
- Be able to apply the discipline of separating evidence from fiction by labeling each reconstruction with its level of evidence and clearly marking AI images as 'reconstruction'
- Ability to understand the responsibility of expressing uncertainty honestly and establishing cultural representation through community participation and away from stereotypes.
Archeology is not just for experts; The past belongs to the entire society. Public archeology (the practice of conveying archaeological knowledge to the public in an understandable, engaging and honest way) operates through museums, exhibitions, digital narratives and reconstructions. In this unit, we will see how AI helps in the production of exhibition text, multilingual access, visualization and reconstruction (recreating the possible state of a structure, object or scene in the past), but why honesty, evidence-fiction distinction and cultural sensitivity must be maintained at every step.
Basic principle: AI is a powerful tool for producing accessible text, translation and storyboarding; but the accuracy of the history described, the separation of evidence from imagination, and the representation of communities are your scientific and ethical responsibility. A museum is obliged to tell the truth to the visitor; Attraction cannot trump accuracy.
Roles of AI in public archaeology
Exhibition and label text. It is difficult to translate expert jargon into language the visitor can understand. AI is fast at converting complex information into a simple outline; But each sentence is linked to the source and confirmed by an expert.
Multilingual access. Museum texts are translated into different languages; AI gives draft translation, expert/native speaker corrects (see unit 6 translation discipline).
Visualization and reconstruction. The possible state of a temple, a city or a piece of clothing is visualized. This is the most delicate area: reconstruction inevitably continues where the evidence ends, with imagination.
Accessibility. Descriptions, sign language drafts, and simplified narratives are produced for the visually impaired.
Tip: Put a “level of evidence” note next to each reconstruction: make it clear what is based on hard evidence, what is based on parallel example, and what is guesswork. Ask the visitor "how much of this do we know?" Answering the question is the duty of honesty of public archaeology.
The ethics of reconstruction: where does the evidence end?
A reconstruction can be beautiful and convincing; but that doesn't make it right. The danger is that the visitor may think the prediction is real. For example, just because a statue appears to be white marble today does not mean it was also white in antiquity — many were painted. AI is prone to producing a “nice” and “expected” image and makes these mistakes:
- Fills in the gaps with confidence: "plausibly" fits details that are not in evidence (colour, trim, roof).
- Slips into cliché: Repeats popular but inaccurate images in the training data (e.g., "ancient" appearances in movies).
- It hides uncertainty: It presents a single definitive image, whereas there are many possible states.
Solution: always present the reconstruction as “a possibility”, with a level of evidence label, and alternatives if possible.
Caution: An AI-generated image may look like a real photo. It is an ethical obligation to make it clear to the visitor that this is a reconstruction (not a photograph of an actual artifact). An unlabeled AI image creates a false “memory” of the past.
Cultural sensitivity and representation
Past societies and their current heirs have a say in how their stories are told. Some objects, sites or human remains may be sacred or sensitive to indigenous and local communities. AI can stereotype a culture, repeat a colonialist view, or misrepresent a community. The narrative is constructed collaboratively and respectfully with the communities involved.
three mini cases
Case 1 — Access expanded. A museum simplified exhibition texts with AI and translated them into 5 languages, making them accessible to children and foreign visitors; Visitor comprehension rate increased with feedback. Each text was verified by the curator and native speaker.
Case 2 — Fabricated detail caught. One team was going to put an AI-generated reconstruction of an ancient city on display; The archaeologist noticed a dome for which there was no evidence in the image and uncertain colors. The image was accompanied by an alternative version with a level of evidence label and an “artist comment” note.
Case 3 — Stereotype corrected. In one digital narrative sketch, AI depicted a community in a stereotypical and reductive manner. The curator and community representatives rewrote the text together; the narrative was adjusted to include the community's own voice.
Four copyable templates
1) Accessible exhibition text:
Your role: museum copy editor. Translate the following expert text into plain language that the general visitor can understand, WITHOUT IMPROVING its accuracy. Explain the jargon. Do not add any information that is NOT in the source; Mark where you are not sure. At the end of the text, indicate which sentence is based on which source. Subject to curator approval.
2) Reconstruction evidence note:
I'm preparing a reconstruction. For the following items (structure, color, trim, roof, furniture), help me classify what level of EVIDENCE each is based on: (a) direct evidence, (b) parallel example, (c) guess/imagination. Clearly mark those that have no evidence as "assumption". Write a draft description note to show to the visitor.
3) Cultural sensitivity audit:
Evaluate the following narrative/exhibition text for cultural representation: is there a risk of stereotyping, reductionism, colonialism, or misrepresentation of a community? Flag potentially sensitive phrases and suggest questions to ask in collaboration with the community. "Correct" text alone; show risks.
4) AI image tagging:
I will use an AI-generated image in an exhibition. Write a draft label/caption text that will not mislead the visitor and clearly state that this is a reconstruction, not a real photograph, and which parts are based on evidence and which are based on interpretation.
Weak prompt / Strong prompt
Weak prompt:
Let's produce a realistic visual of this ancient city and put it in the exhibition.
The AI confidently invents details that have no evidence, slides into cliché, and if it is not labeled, the visitor will think it is real.
Powerful prompt:
For a reconstruction image of this ancient city, use only the evidence I have (plan, find, parallel example). ADD elements for which there is no evidence (colour, roof, trim) or explicitly mark them as "assumption". Also suggest an alternative possible version. Write label text stating that the image is not an actual photo.
The difference: the former produces a misleading "truth"; The second establishes a narrative that is honest, labeled, and has a certain level of evidence.
Explaining uncertainty: public archeology's most difficult skill
The most subtle task of public archeology is to describe uncertainty honestly but engagingly. The visitor wants definitive answers; However, archeology often says, "We don't know for sure, but the most likely is this." A bad narrative disguises this uncertainty and sells false certainty; A good narrative, on the other hand, makes the visitor share in how science works by saying "here is the evidence we have, here is our interpretation, here is what we still don't know." AI is helpful in producing outlines that explain a complex uncertainty in plain language; but the natural tendency of AI is to flatten uncertainty and present a single, definitive story. It is necessary to correct this consciously.
Another powerful tool is to present multiple comments together. If the function of a building (temple or storehouse?) is in dispute, it is both more honest and more interesting to show the visitor two possible interpretations and the evidence on which each is based, rather than imposing a single "truth" on the visitor. This teaches that archeology is not a “repository of answers” but a “reasoning based on evidence.”
Finally, making the source and process visible creates trust. "How do we know this?" An exhibition that answers the question - showing which find, which analysis, which comparison was used - allows the visitor to both learn and trust science. Every text and image produced with AI must pass this transparency criterion.
Tip: When you finish an exhibition text, ask yourself: "What will the visitor be able to distinguish from here as knowing for sure and what as 'possible'?" If everything is presented with the same certainty, you need to put back the uncertainty.
Public archeology task table
Quest
Role of AI
human decision
code of ethics
Exhibition text
Simplification sketch
accuracy, signature
loyalty to source
Translation
draft
Nuance confirmation
Native language confirmation
reconstruction
storyboard
Evidence-fiction distinction
Labeling
Narrative
text draft
representation, voice
community involvement
accessibility
description sketch
Eligibility
accuracy
Common mistakes
- Presenting the reconstruction without labels. The visitor thinks the prediction is true.
- Not separating evidence from imagination. Without a level of evidence note, the narrative is misleading.
- Putting appeal over accuracy. The museum first tells the truth.
- Repeating cultural stereotype. Representation is established through community participation.
- Making the AI visual look like a photograph. Reconstruction must be clearly stated.
In summary
AI in public archaeology; It is a powerful aid in accessible text, multilingual access and storyboard production, bringing history to wider audiences. But honesty is essential: every narrative is linked to the source, every reconstruction is tagged with its level of evidence, evidence is separated from imagination, AI images are clearly marked, and communities have a say in their own representations.
Application task
Prepare both an expert-level description of a find or structure and a simplified visitor text with the "Accessible exhibition text" template. For a reconstruction idea, divide the elements into evidence level with the "Reconstruction evidence note" template and write a label text. Finally, review your text for representation with the “Cultural sensitivity check.”
checklist
- [ ] I linked the exhibition text to the source and left it to the curator's approval.
- [ ] I separated reconstruction elements by level of evidence.
- [ ] I clearly labeled the AI image as a “remake.”
- [ ] I did not put appeal before accuracy.
- [ ] I created cultural representation with community participation and respect.