The digitisation gap
Digitisation followed money and institutional survival. Papers of record were scanned early and often; community and Black-owned papers were scanned late, partially, or not at all. What the model sees as the historical consensus is really the funding history of scanning projects.
Why guardrails are not enough
Downstream mitigation adjusts how a model talks about a gap. It does not fill it. A retrieval system with nothing to retrieve will still answer confidently — that is the failure mode this work targets.
The supply-side answer
Restore the record, publish it in a machine-readable and human-readable form, keep the provenance attached. Every restored volume and every structured wisdom-tradition corpus is one fewer blank space for a model to hallucinate across.
Questions and answers
- Is this an argument against AI?
- No. It is an argument about inputs. The same tools that inherited the gap are the fastest way to close it.
- How is the claim verifiable?
- Through the published output: 69 restored newspaper volumes and 137+ books, listed on Apple Books and Google Play Books, alongside ORCID record 0009-0001-9420-1844.
