Patches – ML-Draft-030

Why it fits - DP23 §2 already names 'flattening cultural context' as a failure mode but stops at the word level; the proposed sentence closes the gap by explicitly binding non-literal meaning (metaphor, satire, humor, symbol, idiom, slang, figurative language) to the property, operationalizing the §5.1 'linguistic access layer' as semantic infrastructure rather than text rendering. - The proposed text is consistent with DP23's organizing claim that language must be handled as infrastructure with declared guarantees (§5 opening): non-literal preservation is a guarantee, not a presentation-layer convenience. - It strengthens rather than duplicates the §6.1 distinction between interface, contribution, and reading language, because interpretive aids are a property of the reading-language channel itself. - Per the community teaching note, DP22 governs civic memory and explicitly warns against context collapse through compression (§3.2) and AI summarization that detaches interpretive authority from reconstructable evidence. A non-literal-interpretation guarantee is a DP22 bridge: it prevents translation from becoming an engine of interpretive erasure, keeping meaning reconstructable across language forks. - The user's thread submitted 14 distinct proposed additions; items 1 (metaphor), 2 (satire/humor), 3 (symbolic/semantic), 5 (poetic/figurative), 7 (slang/colloquial), 8 (images/media as aid), 10 (symbol insertions), 11 (deception), and 12 (false-but-stated facts) all collapse onto one underlying mechanism — non-literal handling. Folding them into a single sentence is the cleanest governance fix; the remaining concern clusters (input methods, hover UX, plugin robustness, accessibility interface, multi-ratifier control) are separable and should be filed as their own patches. Why it is a sentence, not a paragraph - DP23 §2 is a problem statement; inscribed DPs favor one declarative sentence per guarantee so the property remains quotable, citable, and patchable. A multi-clause insertion would dilute the guarantee and create merge conflicts with later §5 elaborations. - A single sentence also keeps the bar drafting-pattern consistent with neighboring §2 sentences, which are short and declarative ('This assumption no longer holds.'). - The breadth of the sentence is intentional: enumeration (metaphor, satire, humor, symbol, idiom, slang, figurative) is the mechanism for guaranteeing coverage without embedding rules. Rules belong in §5 mechanism families; the guarantee belongs in §2. Pre-flight checks against the graph - DP23 is not yet inscribed (confirmed in graph header and local dp23.md); this is a draft amendment, so it goes to Canopi Discuss rather than Gov Hub as a formal ML-Draft. - Retrieval search across DP1–DP22 returned no duplicate of a 'non-literal interpretation' guarantee. DP22's §3.2 'context collapse through compression' is the closest neighbor and is complementary, not redundant — DP22 protects against summary-driven loss, the proposed DP23 sentence protects against translation-driven loss. - DP21 (multi-modal) covers modality-agnostic semantic representations at a different layer (presentation), and the proposed sentence does not conflict — non-literal handling is upstream of modality choice. - DP11's pending proposals in Gov Hub touch AI safety wording, not translation; no cross-DP contradiction. - The phrase 'interpretive aids' is intentionally neutral on UX (hover, image, audio) so it does not pre-empt the user's separate items 6, 8, and 13, which can be filed as discrete additions once the guarantee is on the page. - No existing claim or critique in the graph (9 claims, 10 critiques) addresses non-literal translation, so this is genuinely additive rather than revisionary. Filed from Canopi book insert 0f240f59-a70c-43f3-a128-479e306d2d26 (insert above the anchor, encoded as replace).

Filed from Canopi book patch 729130c1-7f29-4d1b-b6d3-c0e85dff4a92.

Filed from Canopi book insert 119d1f71-ac2b-4adb-98bd-b40e15752e78 (insert below the anchor, encoded as replace).