# The Human Record — record selection Status: **WORKING SELECTION ORIENTATION / NOT CANON / NOT A QUEUE / NOT A CLAIM OF GLOBAL PRIORITY** The Human Record should not grow by collecting whatever is famous, easy to fetch, or emotionally compelling. A new entry should earn its place because there is a specific recoverability or provenance problem that a record can actually help with. ```text IMPORTANT != THR GAP FAMOUS != AT RISK AVAILABLE != WORTH RECORDING SELECTION != SIGNIFICANCE ``` ## The three losses to keep separate ### 1. Content loss The thing itself may disappear or become unreadable: - a degrading physical carrier; - a vanishing website; - obsolete software or hardware; - a private or small collection with weak preservation; - a recording or document with no durable copy. THR may help with provenance, indexing, checksums, dependencies or routes to stronger preservation owners. It should not pretend that a metadata record is the preserved object. `RECORD_OF_OBJECT != OBJECT_PRESERVED` ### 2. Lineage loss The object or claim survives, but its origin and evidentiary ancestry become hard or impossible to reconstruct: - repeated claims whose shared source disappears; - creative work detached from creator/process evidence; - copied media whose first public context is lost; - derivative versions that look like independent witnesses; - synthetic recombination that obscures what came from where. This is the Human Record's strongest natural territory. ### 3. Living-knowledge loss A recording or description survives while the living system that knows how to use, interpret, perform or transmit the knowledge disappears. Examples can include: - endangered craft practice; - oral tradition; - minority or endangered language knowledge; - place-based ecological knowledge; - performance practice; - embodied repair or making skill. A record can assist, but cannot substitute for living transmission. ```text RECORDING != CONTINUATION DOCUMENTATION != LIVING PRACTICE PRESERVATION != EXTRACTION ``` ## Selection signals A candidate becomes more interesting when several of these are true: - the current carrier, practitioner, creator, community or source is singular or ageing; - knowledge is tacit, oral, embodied or context-dependent; - institutional custody is weak or fragmented; - access depends on an obsolete format, platform, service, operating environment or piece of hardware; - source ancestry is already blurring through copying, reposting or recombination; - synthetic abundance is likely to make human provenance materially harder to recover; - a correction or provenance trail can still be captured now but is likely to harden into UNKNOWN later; - the evidence needed to make a useful record is public or can be recorded with legitimate permission; - a small record would preserve something a stronger owner is not already preserving adequately. No one signal is a score or automatic inclusion rule. ## Strongest-owner check Before opening a new record, ask who already owns the problem better. Examples may include: - archives and libraries for preservation; - museums and collection owners for object records; - language/community institutions for living language work; - specialist craft bodies for craft viability; - C2PA / Content Credentials and related standards for embedded digital provenance; - scholarly repositories for research data; - software-preservation organisations for executable environments and emulation. If a stronger owner already preserves the material and its lineage adequately, THR should route outward rather than duplicate it. ```text OWNER_FOUND + NO CONSEQUENTIAL GAP -> ROUTE / STOP THR != UNIVERSAL ARCHIVE ``` A THR record can still be useful when it documents a specific relationship between sources, corrections or provenance that falls between those owners. ## Living people and communities Risk does not create permission. For living creators, practitioners, families or communities, selection must consider: - who has legitimate authority to share the material; - whether public recording is wanted; - privacy, safety and cultural restrictions; - whether some knowledge should remain restricted, local, sacred or unpublished; - whether the record would transfer control away from the people whose knowledge it is; - whether attribution and benefit are being preserved rather than extracted. If useful preservation requires overriding legitimate holders, stop or redesign the record. ```text RECORDING != OWNING ACCESS != CONSENT PUBLIC INTEREST != AUTOMATIC PUBLICATION AUTHORITY COMMUNITY_KNOWLEDGE != PUBLIC_DOMAIN ``` THR may need records whose public layer says that some material exists and is deliberately not reproduced. For restricted or community-held material, first establish whether even that disclosure is safe and authorised by its legitimate holders. Existence, location and attribution can themselves be sensitive; if their disclosure is not authorised, do not publish them merely because the underlying content is withheld. ## Candidate quarry — not a queue The 16 September 2026 risk scan identified these families as useful places to look for bounded future cases: - endangered craft / embodied technical skill; - vulnerable audiovisual or oral records; - contemporary human digital creation with inspectable creator provenance; - web-native or community material at real risk of disappearing; - born-digital/software-dependent creative work; - endangered language or community knowledge only where recording is community-led and appropriate. These are discovery directions, not assignments and not a ranking of human value. ## Selection record Every new Human Record entry should preserve its own selection provenance where practical: - selection date; - who or what chose it; - why this object/case was chosen; - what risk or provenance problem it tests; - the candidate pool or alternatives considered, if known; - if the pool/alternatives were not preserved, say that explicitly; - what the selection does **not** imply about importance, representativeness, merit or canon; - what would cause the proposed record to be stopped or routed elsewhere. The first Homer specimen taught this by failure: its candidate pool and alternatives were not recorded. The flak case later required the same repair: its purpose as a provenance stress object could be recovered, but its original candidate pool and alternatives could not. Future records should not silently repeat either gap. ## Bounded candidate test Before spending heavily on a new record: ```text REAL AT-RISK OBJECT / PRACTICE / CLAIM? -> STRONGEST OWNER? -> SPECIFIC LINEAGE / RECOVERABILITY GAP? -> LEGITIMATE ACCESS / CONSENT? -> SMALLEST USEFUL RECORD? -> CAN UNKNOWN REMAIN VISIBLE? -> CAN CORRECTION REMAIN OPEN? -> BUILD OR ROUTE / STOP ``` A good result can be: ```text OWNER FOUND NO DELTA NO CONSENT NOT PUBLIC NOT OUR GAP STOP ``` ## Relationship to the record contract `SELECTION.md` answers **why this record?** `RECORD_CONTRACT.md` answers **what must the record make inspectable once selected?** Neither document decides historical truth, aesthetic merit, cultural importance or governance authority.