← Augmented Philippine Intelligence

Agent Architectures for Cultural Preservation

#agents #cultural-preservation #architecture #oral-traditions

The Urgency

The Philippines has over 180 living languages. Ethnologue classifies 28 of them as “in trouble” and 11 as “dying.” When a language dies, it takes with it an entire epistemology — ways of naming plants, describing weather, narrating history, and encoding law that exist nowhere else.

Traditional preservation methods — field recordings, dictionaries, academic papers — are valuable but insufficient. They capture artifacts, not systems. A recording of a Tboli creation myth preserves the words but not the web of cultural knowledge that gives those words meaning.

Multi-agent systems offer a different approach: preserve the relationships, not just the records.

The Constellation

The API cultural preservation stack deploys specialized agents in concert:

Agent-Transcriber

Processes audio recordings of oral traditions. Uses multilingual speech-to-text fine-tuned on Philippine languages, then annotates the transcript with:

  • Speaker identification and kinship context
  • Geographic and temporal references
  • Intertextual links to other narratives
  • Linguistic features (code-switching, archaisms, loanwords)

Agent-Ethnobotanist

Maps indigenous plant knowledge from collected narratives. For each plant reference, the agent records:

  • Local name(s) across languages
  • Taxonomic identification where possible
  • Documented uses (food, medicine, material, ritual)
  • Associated narratives and taboos

Agent-Cartographer

Builds spatial models of cultural geography — not the colonial maps that imposed new names, but the indigenous mental maps embedded in oral traditions:

  • Named places and their significance
  • Travel routes and trade paths
  • Sacred spaces and boundaries
  • Seasonal movement patterns

Agent-Curator

Catalogs and cross-references material culture: textiles, pottery, metalwork, musical instruments. Links artifacts to the communities that produced them, the techniques used, and the narratives that explain them.

The Coordination Layer

Individual agents produce domain-specific knowledge. The coordination layer produces understanding:

  • Conflict resolution: When Agent-Ethnobotanist and Agent-Transcriber disagree on a plant identification, the coordinator flags the discrepancy for human review
  • Cross-domain linking: When a textile pattern (Agent-Curator) matches a creation narrative (Agent-Transcriber), the coordinator proposes a semantic connection
  • Temporal reasoning: The coordinator maintains a timeline, placing cultural elements in historical context
  • Gap detection: The coordinator identifies domains where knowledge is thin and prioritizes future collection efforts

Ethical Architecture

Cultural preservation AI must be built with ethical guardrails:

Consent and ownership. Indigenous communities own their cultural knowledge. Agents process material only with explicit community consent, and outputs are governed by community data sovereignty agreements.

Access control. Some knowledge is sacred and restricted. The system supports tiered access — some knowledge is public, some is community-only, some is restricted to initiated individuals.

Benefit sharing. Any commercial application of culturally-derived insights must flow benefits back to the source communities under mutually agreed terms.

Reversibility. Communities can withdraw consent and have their cultural data removed from the system entirely.

Starting Points

The first deployment targets three communities:

  1. Tboli of South Cotabato — Preserving t’nalak weaving traditions and the dream narratives that generate designs
  2. Sama-Bajau of Tawi-Tawi — Mapping maritime knowledge systems and boat-building techniques
  3. Ivatan of Batanes — Documenting vakul culture and typhoon-adapted architecture

Each deployment is co-designed with community leaders, using their priorities to determine what agents are deployed and what knowledge is processed.

The Objective

The goal is not to digitize culture. It is to build systems that help living communities maintain and transmit their knowledge to future generations — in their own languages, on their own terms, with technology that serves rather than extracts.

References: Ethnologue: Languages of the World, 27th ed., ed. David M. Eberhard, Gary F. Simons, and Charles D. Fennig (SIL International, 2024) — Philippine language vitality assessments; UNESCO, Atlas of the World’s Languages in Danger (2010); S. Lily Mendoza, Between the Homeland and the Diaspora: The Politics of Theorizing Filipino and Filipino American Identities (2002); Manolete Mora, Myth, Mimesis and Magic in the Music of the T’boli, Philippines (2005); Harry Arlo Nimmo, The Sea People of Sulu: A Study of Social Change in the Philippines (1972); Florentino Hornedo, Culture and Community in the Philippine Fiesta and Other Celebrations (2000). Technical: Michael Wooldridge, An Introduction to MultiAgent Systems (Wiley, 2nd ed., 2009); Stuart Russell and Peter Norvig, Artificial Intelligence: A Modern Approach (4th ed., 2020); CARE Principles for Indigenous Data Governance (Global Indigenous Data Alliance, 2019); United Nations Declaration on the Rights of Indigenous Peoples (2007), Articles 11, 15, 31.