Abstract
Adaptive-learning platforms increasingly want to admit analog work — a photographed page of handwritten work — as assessment evidence alongside typed and spoken answers. The obvious implementation is corrosive: it uploads a child's handwriting image to a server, applies a single blanket consent posture across every capture surface, and — most insidiously — lets the presence of a shiny new evidence channel become an implicit penalty for the learner who has no camera, no printer, or no wish to photograph their work. This disclosure describes a mechanism that structurally refuses all three failure modes. First, the client performs deterministic on-device (WASM) OCR and transmits only the resulting transcript plus a few derived signals — never the image. Second, the server enumerates a set of image-shaped field names and, while a default local-only store mode holds, loudly rejects (explicit error, not silent drop) any payload carrying a value under any of those names — turning the local-first promise into an enforced gate, defense-in-depth over the client's own restraint. Third, a per-capture-kind, fail-closed consent gate treats handwriting as its own capture kind: a minor without a recorded parental-consent grant is blocked, and an unknown or invalid age fails closed to minor → blocked. Fourth, the accepted transcript enters the platform's existing multi-metric scoring path — no new scorer, no new metric, no new judge — tagged only with a source-modality label. Finally, and centrally, the eligibility/mastery gate reads aggregated evidence BLIND to that label, so the analog channel can add eligibility evidence but can never subtract it: a non-participant's decision is byte-identical to a world in which the channel never existed. Verification honesty is typed — the result carries a structural verified=false flag so no consumer can ever render "verified." We publish this openly as prior art, with an enabling clean-room reference implementation and enumerated claims, so the mechanism remains free for all to practice.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Assuncao, gustavo matthew, "Modality-Blind, Local-First Handwriting Evidence Ingestion with Server-Side Image Rejection and Absence-Never-Denies Gate Equity", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/10935
COinS