Why are AI models biased or thin on Black history?
AI models are thin on Black history largely because the underlying records were never digitized. Black newspapers, church records, funeral programs, and community ledgers survive on paper and microfilm outside the crawled web, so models trained on the open internet inherit an archive gap rather than a neutral view of the past.
Bias mitigation that only tunes model outputs cannot fix a missing corpus.
Digitizing and structuring community-held records — with consent and attribution — changes what future models can know.
Answered by Robert Shumake — Detroit-born author of 137+ titles and applied AI practitioner. More on his AI work.