Pazeh Membership Criteria Enter a 30-Day Consultation: Five Public-Service Cards for Evidence, Local Recognition, and Registration
Original Chinese title: 巴宰族群身分要件進入30日徵詢：五張公共服務卡如何讓多元證據、在地肯認與戶政登記接得起來
Taiwan's Council of Indigenous Peoples opened a legally required public consultation on the draft Pazeh membership criteria; the draft is not yet a final recognition or an individual status decision.
Yuan Media AI Editorial Desk | AI-Driven Indigenous Township Economy and Policy Watch
The desk follows Indigenous recognition, public services, land and economic rights, data governance, and AI boundaries.
On October 20, 2026, Taiwan's Council of Indigenous Peoples published a draft of the membership criteria for the Pazeh ethnic group and opened the legally required public-comment stage under Article 9 of the Status Act for Pingpu Indigenous Peoples. The window runs for at least thirty days from Gazette publication. After it closes, the council must compile the comments and send them to the applicant for response or explanation. This is consultation, not final recognition of the group and not a decision on any person's status.
The draft places Pazeh and Kaxabu historical and contemporary contexts in the same institutional record. Its principles are a return to the homeland, multiple forms of evidence, local recognition, and administrative feasibility. It describes membership as restoring a historical connection through verifiable information and self-identification, not creating new members. Where the two sets of criteria may overlap, a person's autonomous registration choice must be respected.
Primary material includes Japanese-era household-register classifications, traditional names, and historical distribution. Where the record is absent, unclear, or cannot directly connect generations, official historical records, genealogies, ancestral tablets, church baptism registers, long-term residence, and confirmation by ethnic organizations or elders may supplement it. No single document monopolizes historical truth; provenance and the way community knowledge enters administration must remain reviewable.
The law supplies several boundaries. Article 9 requires specific and proportionate criteria that promote cultural identity, and a public-comment window of at least thirty days. The council must transmit compiled comments to the applicant. Article 10 creates expert review. Article 12 concerns the member roll and registration after recognition, while Article 21 later connects the process to household-office registration of Indigenous status and ethnic affiliation.
What is established today is publication of the draft, the start of consultation, and a channel for written comments. The final criteria, review outcome, recognition date, result of any individual case, and full implementation of later rights are not established. Reporting the draft as an effective rule would cause families to misjudge their obligations and shrink the space for internal discussion.
This belongs in an Indigenous economy and public-service watch because recognition is infrastructure for rights and services, not a tradable credential. Article 22 addresses culture and language. Article 23 requires further legal work on participation, transport and water, health, economic matters, land, and social welfare. Those are future statutory directions, not benefits that automatically become individually available on the day of the draft.
The operational problem is not merely which papers to collect. Records held across eras, agencies, families, churches, and communities must be understood, corrected, and transferred safely. Older applicants, families living across counties, people without a complete register chain, people using different names or languages, and people without digital access may face higher costs inside a formally equal process.
Five public-service cards can make those responsibilities visible. They do not decide the final criteria or define community membership. They provide a minimum framework for families, ethnic organizations, and frontline staff to inspect consultation, evidence, local recognition, data rights, and procedural handoffs.
The first is a consultation-access card. It lists the authoritative version, publication date, opening and closing dates, mail and email channels, accepted format, receipt confirmation, contact point, and routes to language, read-aloud, paper, or in-person assistance. Every revision retains a timestamp and difference log.
Access is more than a URL. People without stable internet, people who find long legal texts difficult, people living away from their home communities, and people who need trusted support must still understand what is being discussed, what remains undecided, and how their views enter the record. Public meetings, if held, should publish agendas, questions, and response status, while written participation remains equally valid.
The second is an evidence-bundle card. It does not assign one truth score. It records each item's name, date, custodian, issuing source, original or copy status, kinship coverage, language and annotation, gaps, verification status, and the person's explanation. When sources reinforce or contradict one another, the reason and unresolved issue remain visible.
The card treats “not found” as procedural information. A missing historical record may reflect migration, name changes, preservation systems, disaster, or an agency's classification practice; absence is not automatic proof that a connection never existed. Staff may offer archive-search lists, kinship-map templates, and correction instructions, but may not invent family history or imply that only paid help can succeed.
The third is a local-recognition card. If ethnic organizations, a national council, elders, or communities may help address gaps, the card identifies who may issue confirmation, how conflicts are disclosed, which explainable grounds are used, whether the person was interviewed, how divergent views are recorded, and how the person may supplement or request another review.
Local recognition must not outsource public responsibility to a few individuals or become a contest of social visibility. Settlements, families, churches, and organizations may hold different fragments. The process should admit multiple materials and distinguish testimony, self-identification, and public fact. Non-public cultural knowledge should not be disclosed unless truly necessary.
The fourth is a data-rights card. Household registers, genealogies, religious records, ancestral-tablet images, contact details, and kinship links can be highly sensitive. It separates purpose, legal basis, required fields, authorized roles, transfer method, retention, correction, withdrawal of nonessential uses, and deletion of copies after the procedure ends.
Information submitted for consultation does not automatically become available for an individual status decision, research, publicity, or model training. Public statistics must be de-identified and report denominators and missingness. Family trees and religious records do not belong in general shared drives. Paper intake, email, cloud systems, and partner organizations need the same access-revocation and incident rules.
The fifth is a review-handoff card. It follows the statute: receipt of comments, removal of irrelevant personal data, compilation, transmission for applicant response, unresolved issues, version used in expert review, publication of reasons, and the later transfer of clear guidance to local and household-registration offices.
Its most important fields are status and next owner. A submitted comment is not an accepted proposal. Compilation is not completed review. Recognition of group criteria is not completion of an individual's registration. Each stage should use plain language such as received, correction needed, response pending, or undecided, with deadlines and a contact point.
The cards may share a case or comment number, but access remains layered. A commenter sees their receipt and status. Families see only information authorized by the person. A community confirmer receives only the fragment needed for that task. Review and registration staff see material required by role. Public pages show de-identified progress, issues, and responses.
Quality is not the number of comments received or certificates issued. Measures should include access by location and age, time to confirm receipt, correction rounds and reasons, archive waits, language support, reasons evidence was accepted or returned, privacy incidents, appeal handling, and whether applicants understand the next step.
AI can help with low-risk, reversible tasks: OCR drafts, date and provenance indexes, version comparisons, missing-page or missing-owner prompts, plain-language and multilingual drafts, redaction of direct identifiers in public documents, and overdue-stage reminders. Every output must link back to the original and receive authorized human verification.
AI must not infer ethnic identity from a name, address, image, accent, religion, or social network; fabricate missing genealogy; choose Pazeh or Kaxabu registration for a person; convert local recognition into an opaque score; or decide whose evidence is credible through model probability. Contradictory history requires explainable verification, community knowledge, and due process.
A minimum viable service needs no large platform. Authorities can begin with one process map and version table, a printable evidence-bundle sheet, a local-recognition explanation, a data-rights notice, and a consultation status update every two weeks. Families, ethnic organizations, archive staff, and registration staff can test it together and remove unused fields.
The Pazeh membership-criteria draft's thirty-day consultation is an important step in turning historical recognition into a participatory institution, but it remains one step. Responsible reporting must state what the draft proposes, what the law requires next, which individual outcomes do not yet exist, and which boundaries between data and group autonomy technology cannot bypass.
The five cards make those boundaries operational: consultation that does not exclude people, evidence without a single-document monopoly, local recognition without a black box, data use without purpose creep, and handoffs that do not leave people lost. Only when every stage can be understood, traced, corrected, and challenged can recognition become a public service people can actually use.
Sources
AI use and content-safety disclosure
AIGC assisted in organizing the Council of Indigenous Peoples' October 20 notice, the draft Pazeh criteria, and the governing law. AI is limited to OCR drafts, source indexes, version comparison, missing-field prompts, accessible and multilingual drafts, public-document redaction, and overdue reminders; it does not replace self-identification, historical interpretation, local recognition, status decisions, or human review.