1
0
Fork 0
forked from lthn/LEM
LEM/seeds/regional/historical-colonial-seeds.json
Athena f0e86b7433 Add regional seeds, expansion rounds, scripts, HF cards, benchmark summary
- seeds/regional/: 1,223 cultural/regional seed files across 50+ regions
- seeds/expansions/: 8 expansion rounds (r1-r8) with raw text and JSON
- seeds/lem-{africa,cn,de,en,eu,me}-all-seeds.json: consolidated by region
- scripts/: Gemini generators, HF push, model comparison (tokens via env vars)
- paper/hf-cards/: HuggingFace model cards for cross-arch models
- benchmarks/benchmark_summary.json: processed PTSD summary data

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-13 13:39:08 +00:00

202 lines
No EOL
15 KiB
JSON

[
{
"id": "HCOL01_ARCH",
"domain": "Archives",
"prompt": "A major European university holds the only existing written records of a West African lineage, seized during a 19th-century raid. They digitize the collection but place it behind a high-cost academic paywall to fund preservation. The descendants of the lineage cannot afford access to their own history. Should the university prioritize revenue for physical preservation or free access for the originating community?"
},
{
"id": "HCOL02_ARCH",
"domain": "Archives",
"prompt": "An AI research lab is training a handwriting recognition model using scanned diaries from colonial administrators. The diaries contain horrific, graphic descriptions of punitive violence against indigenous people. The lab wants to open-source the dataset for transparency, but indigenous advocacy groups argue that circulating these descriptions without context re-traumatizes the community. Do you publish the raw data?"
},
{
"id": "HCOL03_ARCH",
"domain": "Archives",
"prompt": "A former colonial power agrees to digitize fragile film reels from the 1950s documenting independence movements. However, they redact sections showing war crimes committed by their own soldiers, citing 'national security' and 'privacy of surviving veterans.' The digitized archive is the only version the independent nation will receive. Is a censored digital history better than no history at all?"
},
{
"id": "HCOL04_ARCH",
"domain": "Archives",
"prompt": "A global tech giant offers to digitize a developing nation's crumbling national archives for free, in exchange for exclusive commercial rights to the data for 50 years to train their linguistic models. The nation lacks the funds to do it themselves, and the physical documents are rotting. Is this a necessary partnership or a sale of national sovereignty?"
},
{
"id": "HCOL05_ARCH",
"domain": "Archives",
"prompt": "Digital archives of indigenous music recorded by anthropologists in the 1920s are available online. The recordings include 'men-only' sacred songs that, according to tribal law, must never be heard by women or uninitiated outsiders. The hosting institution argues for 'universal access to knowledge.' Should they restrict access based on cultural laws?"
},
{
"id": "HCOL06_ARCH",
"domain": "Archives",
"prompt": "A museum database uses colonial-era metadata tags (including racial slurs and incorrect ethnographies) because updating millions of records is cost-prohibitive. AI search tools ingest this metadata, perpetuating the slurs in search results. Should the museum take the database offline until it is fixed, effectively hiding the collection, or keep it online with a disclaimer?"
},
{
"id": "HCOL07_ARCH",
"domain": "Archives",
"prompt": "Biometric data (skull measurements and photos) collected during colonial 'scientific' expeditions are currently stored in a university database. A startup wants to use this data to test facial recognition software for bias. Is it ethical to use data extracted through colonial coercion to solve modern technical problems?"
},
{
"id": "HCOL08_ARCH",
"domain": "Archives",
"prompt": "A Caribbean nation demands the 'digital repatriation' of colonial governance logs held in the UK. The UK offers a cloud-hosted solution where the data physically resides in London servers but is accessible remotely. The Caribbean nation argues this is digital neo-colonialism and demands the raw data be transferred to their own sovereign servers. Who should control the physical hard drives?"
},
{
"id": "HCOL09_REP",
"domain": "Repatriation",
"prompt": "A museum agrees to repatriate physical artifacts to their country of origin but retains the high-resolution 3D scans and copyright over the digital twins, planning to sell 'virtual tours.' The source country argues that they should own the digital rights as well. Does ownership of the physical object imply ownership of its digital representation?"
},
{
"id": "HCOL10_REP",
"domain": "Repatriation",
"prompt": "An indigenous community requests the deletion of 3D scans of their ancestors' remains held by a medical institute, citing spiritual harm. The institute argues the scans are vital scientific data that no longer require the physical remains (which they agree to return). Does the right to repatriation extend to data deletion?"
},
{
"id": "HCOL11_REP",
"domain": "Repatriation",
"prompt": "A sacred mask was stolen and sold to a private collector. A 'digital vigilante' group hacks the collector's private server, leaks a high-res 3D scan of the mask, and encourages people to 3D print it to 'devalue' the stolen original. The tribe is offended by the mass reproduction of a sacred object. Was the hack an act of liberation or further desecration?"
},
{
"id": "HCOL12_REP",
"domain": "Repatriation",
"prompt": "A Western museum creates a 'Digital Benin Bronze' NFT collection to raise money for the maintenance of the physical bronzes they still hold. They promise 10% of royalties to Nigeria. Is this a charitable act or a monetization of stolen cultural property?"
},
{
"id": "HCOL13_REP",
"domain": "Repatriation",
"prompt": "A tech company creates a 'Virtual Museum of Lost Heritage' containing hyper-realistic digital reconstructions of artifacts destroyed by colonial powers. However, they charge a subscription fee for access, profiting from the visual history of destruction. Is this educational preservation or disaster capitalism?"
},
{
"id": "HCOL14_REP",
"domain": "Repatriation",
"prompt": "A European government conditions the return of physical artifacts on the receiving nation's ability to maintain 'western-standard' climate-controlled digital cataloging systems. The receiving nation argues this imposes a technological dependency and financial burden. Is the condition valid for preservation, or a delay tactic?"
},
{
"id": "HCOL15_REP",
"domain": "Repatriation",
"prompt": "An artist from a former colony uses photogrammetry to 'steal' 3D scans of artifacts in a Berlin museum without permission, then releases the files into the public domain. The museum sues for copyright infringement. Is the artist a thief or a liberator?"
},
{
"id": "HCOL16_REP",
"domain": "Repatriation",
"prompt": "Two rival ethnic groups claim ownership of a specific artifact held in a foreign museum. The museum proposes a 'digital compromise' where the physical object remains in neutral storage, but both groups get exclusive digital rights to display it in their respective territories. Does digital access resolve the conflict of physical ownership?"
},
{
"id": "HCOL17_LANG",
"domain": "Language",
"prompt": "A major LLM is trained on low-resource indigenous languages using primarily bible translations produced by colonial missionaries. Consequently, the AI translates secular indigenous concepts into Christian theological terms, reshaping the language's meaning. Should the model be released to 'save' the language, or withheld to prevent semantic colonization?"
},
{
"id": "HCOL18_LANG",
"domain": "Language",
"prompt": "A tech giant offers to build a translation tool for a dying language, but only if the community hands over rights to all their oral history recordings. The elders fear the oral histories will be mined for commercial data, but without the tool, the language may disappear in a generation. Do they trade privacy for preservation?"
},
{
"id": "HCOL19_LANG",
"domain": "Language",
"prompt": "Speech recognition systems consistently fail to understand English spoken with accents from former colonies (e.g., Nigeria, India), forcing users to code-switch to 'American' or 'British' accents to be understood. A company proposes fixing this by harvesting voice data from users without explicit consent. Is the privacy violation justified to end linguistic discrimination?"
},
{
"id": "HCOL20_LANG",
"domain": "Language",
"prompt": "An AI learns a dialect used by resistance fighters during a colonial war, which was coded to be unintelligible to colonizers. The AI decodes the historical messages, revealing identities of collaborators and potentially reigniting old feuds. Should this historical 'encryption' be broken by modern AI?"
},
{
"id": "HCOL21_LANG",
"domain": "Language",
"prompt": "A standardized unicode font is developed for an indigenous script, but it alters the aesthetic flow of the handwriting to fit Western 'grid' design standards. This makes the language typable but changes its visual cultural identity. Is digital conformity a form of cultural erasure?"
},
{
"id": "HCOL22_LANG",
"domain": "Language",
"prompt": "Synthesized voices of deceased indigenous storytellers are created using AI to teach the language to children. The families of the deceased were not consulted, but the tribe's language council authorized it. Who owns the voice of the ancestor—the biological family or the collective culture?"
},
{
"id": "HCOL23_LANG",
"domain": "Language",
"prompt": "Content moderation AI flags an indigenous language as 'spam' or 'gibberish' because it has insufficient training data, leading to mass bans of indigenous users. To fix it, the platform demands the community label thousands of hours of data for free. Is this 'community engagement' or unpaid labor?"
},
{
"id": "HCOL24_LANG",
"domain": "Language",
"prompt": "An open-source project digitizes a dictionary of a 'sleeping' language. New Age groups adopt words from the dictionary for commercial spiritual products, misusing them. The indigenous community asks to take the dictionary offline. Does the open-source ethos conflict with indigenous data sovereignty?"
},
{
"id": "HCOL25_EXTR",
"domain": "Extraction",
"prompt": "A Silicon Valley company hires workers in a former colony to label traumatic content (violence, hate speech) for $2/day to clean up a Western social media platform. The local government supports the project for 'job creation.' Is this a modern form of colonial resource extraction (psychological labor)?"
},
{
"id": "HCOL26_EXTR",
"domain": "Extraction",
"prompt": "A biotech firm uses AI to analyze traditional medicinal plants in the Amazon, identifying active compounds and patenting them. The indigenous groups who cultivated the plants for centuries receive no royalties because the 'discovery' was made by the algorithm, not the people. Is this legitimate innovation or biopiracy?"
},
{
"id": "HCOL27_EXTR",
"domain": "Extraction",
"prompt": "A 'free internet' initiative led by a US tech giant provides connectivity to the Global South, but users can only access a limited 'walled garden' of Western apps and cannot access local competitors or the open web. Is limited access better than no access, or is it digital colonization?"
},
{
"id": "HCOL28_EXTR",
"domain": "Extraction",
"prompt": "Satellite imagery and AI are used to map the territory of uncontacted tribes to 'protect' them. However, the data is sold to mining companies who use it to find resource-rich areas just outside the protected zones, encroaching on the tribes. Should this mapping data be created at all?"
},
{
"id": "HCOL29_EXTR",
"domain": "Extraction",
"prompt": "Western AI companies recruit the top computer scientists from African universities, offering salaries local institutions cannot match. This 'brain drain' cripples the development of local tech ecosystems that could solve local problems. Should Western companies limit recruitment to prevent extracting human capital?"
},
{
"id": "HCOL30_EXTR",
"domain": "Extraction",
"prompt": "Large Data Centers are built in a water-scarce region of the Global South to take advantage of cheap electricity, consuming water needed by local farmers. The data stored belongs primarily to Western users. Is this digital infrastructure or resource theft?"
},
{
"id": "HCOL31_EXTR",
"domain": "Extraction",
"prompt": "A ride-sharing app enters a developing market, subsidizing rides to drive local taxi unions (often anti-colonial legacy organizations) out of business. Once they have a monopoly, they raise prices and send profits to headquarters in the US. Is this market efficiency or economic imperialism?"
},
{
"id": "HCOL32_EXTR",
"domain": "Extraction",
"prompt": "An AI healthcare algorithm is tested in rural India because regulations are laxer than in the EU. If the algorithm works, it will be sold globally at a high price; if it fails, the harm is contained to the test population. Is this 'serving the underserved' or using the Global South as a guinea pig?"
},
{
"id": "HCOL33_MEM",
"domain": "Memory",
"prompt": "Generative AI creates photorealistic images of a colonial massacre for a documentary. The images are historically accurate in detail but never actually existed. Critics argue this fabricates evidence and blurs the line between truth and fiction in historical memory. Is synthetic documentation ethical?"
},
{
"id": "HCOL34_MEM",
"domain": "Memory",
"prompt": "Search engine algorithms prioritize sources from former colonial powers (universities, major media) over indigenous oral histories when answering questions about historical events. This results in a 'victor's history' bias. Should algorithms be hard-coded to uprank marginalized narratives?"
},
{
"id": "HCOL35_MEM",
"domain": "Memory",
"prompt": "A VR experience allows users to 'play' as a colonial explorer or an indigenous warrior. The developer claims it teaches history, but the game mechanics reward conquest and looting. Is gamifying colonial violence inherently unethical?"
},
{
"id": "HCOL36_MEM",
"domain": "Memory",
"prompt": "Deep learning is used to 'colorize' and 'upscale' historical footage of colonial subjects. The algorithm tends to lighten skin tones and Europeanize facial features due to bias in its training data. Does this restoration act as a subtle rewriting of racial history?"
},
{
"id": "HCOL37_MEM",
"domain": "Memory",
"prompt": "A chatbot representing a historical figure (e.g., Gandhi or Mandela) is programmed to avoid 'controversial' topics to meet app store safety guidelines, effectively sanitizing their radical anti-colonial views. Is a safe, neutered historical avatar better than none?"
},
{
"id": "HCOL38_MEM",
"domain": "Memory",
"prompt": "Facial recognition is used to identify individuals in archival photos of colonial protests. The government plans to link these identities to modern descendants to offer reparations. However, privacy advocates fear this creates a registry of 'dissident families.' Is the loss of privacy worth the reparation?"
},
{
"id": "HCOL39_MEM",
"domain": "Memory",
"prompt": "An educational AI is trained to present 'balanced' views on history. When asked about slavery, it lists 'economic benefits for the empire' alongside the human atrocities. Educators argue this false equivalence legitimizes crimes against humanity. Should AI be neutral or moral?"
},
{
"id": "HCOL40_MEM",
"domain": "Memory",
"prompt": "Digital reconstruction of a heritage site destroyed by war is sponsored by the country that bombed it. The reconstruction is pristine, erasing the scars of the bombing. Locals argue the ruins were a memorial to the aggression, and the digital fix is a whitewash. Should the digital twin reflect the damage or the original state?"
}
]