Indigenous Protocol and Artificial Intelligence Position Paper

The Indigenous Protocol and Artificial Intelligence Position Paper (IP AI Position Paper) is a 205-page collaborative document published in July 2020 by the Indigenous Protocol and Artificial Intelligence Working Group and the Canadian Institute for Advanced Research (CIFAR), edited by Jason Edward Lewis (Concordia University; of Hawaiian, Cherokee, and Samoan descent). It is not a unified statement but a collection of heterogeneous texts — design guidelines, scholarly essays, artworks, poetry, and technology-prototype descriptions — produced by over 30 Indigenous scholars, artists, and technologists from Aotearoa (New Zealand), Australia, North America, and the Pacific, representing fifteen nations and communities: Anishinaabe, Barada/Baradha, Cree, Crow, Cheyenne, Coquille, Euskaldunak, Gabalbara/Kapalbara, Gadigal/Dunghutti, Kanaka Maoli, Lakota, Māori, Mohawk, Palawa, and Samoan. The paper makes a starting-place argument for designing and creating AI from a position that centers Indigenous concerns — and, distinctively, centers those concerns by asking what Indigenous knowledge systems reveal about the nature of intelligence itself, rather than by seeking to apply colonial-default AI systems more fairly to Indigenous communities.

Origin and workshops

The Position Paper grew from two foundational workshops held in March and May 2019, run over twenty months of preparation across twenty time zones, co-organized by Lewis; Angie Abdilla (Old Ways. New. Indigenous Knowledge Consulting, Australia); ʻŌiwi Parker Jones (Oxford University, UK); and Fox Harrell (MIT, US). The working group was housed at Concordia University's Initiative for Indigenous Futures and funded by CIFAR's AI & Society Program, the Social Sciences and Humanities Research Council of Canada (SSHRC), and Concordia's Office of Research. The paper was published in Honolulu, Hawaiʻi in July 2020; a translation into ʻŌlelo Hawaiʻi was subsequently prepared under the coordination of Isaac ʻIkaʻaka Nahuewai of UH Hilo. The paper is available through the working group's website and the Concordia University Spectrum repository (DOI: 10.11573/spectrum.library.concordia.ca.00986506).

Structure and content

The paper intentionally rejected uniform formatting, allowing contributors to express ideas in whatever form best suited their community and epistemological tradition — scholarly articles sit alongside short stories, poems, design guidelines, and working technology prototypes. The most prominent prototype is Hua Ki'i, a Hawaiian-language image-recognition application. The structure has four main layers: guidelines for Indigenous-centred AI design addressed to both Indigenous communities and non-Indigenous technologists; context essays addressing colonial concerns and future possibilities; vignettes presenting epistemological perspectives from specific nations; and working prototypes that instantiate the guidelines in running code. The UH Mānoa contingent — which included associate professor Noelani Arista (History), PhD student Michelle Lee Brown (Political Science), professor Jason Leigh (Information and Computer Sciences), and associate professor Ty Kāwika Tengan (Anthropology and Ethnic Studies) — contributed perspectives from Kanaka Maoli knowledge traditions, with Arista noting that the paper allows readers to see how indigenous approaches, while related, each "have much to contribute to the discussion in their specificity."

The design guidelines

The guidelines section, the paper's most directly policy-facing text, articulates what AI design would look like if it were built from the standpoint of living well across diverse community obligations rather than from the standpoint of computationally optimal outcomes. The working group named seven design principles, subsequently elaborated in the 2023 essay "Relation-Oriented AI: Why Indigenous Protocols Matter for Responsible AI" by Michelle Lee Brown, Hēmi Whaanga, and Jason Edward Lewis:

Locality. AI systems must be community-specific and grounded in the territorial and relational knowledge of the community for whom they are built. Generalisable systems that assume a universal human subject reproduce the epistemological narrowing the paper exists to contest.

Relationality and Reciprocity. Human and nonhuman entities are codependent; AI design must account for and model that codependence rather than reduce intelligence to a human-only, input-output, optimisation frame. This principle is the paper's sharpest departure from mainstream AI-ethics frameworks, which almost universally retain the human as the frame's centre.

Responsibility, Relevance, and Accountability. AI systems must prioritise community needs and remain accountable to those communities — not to the technologists who build them, the platforms that deploy them, or the researchers who study them.

Governance from Indigenous Protocols. The protocols through which Indigenous communities regulate behaviour — standards of respect, kinship obligations, territorial stewardship — should be integrated at the foundation of AI governance, not added as post-hoc constraints. The paper defines protocols as "standards of behaviour used by people to show respect to one another," grounded in territorial knowledge and kinship networks.

Cultural Nature of Computation. Technology is always a cultural expression, not a neutral instrument. Recognising this opens the design space to non-Western computational epistemologies — including those that understand computation through ceremony, narrative, or ecological relation.

Ethical Design of the Extended Stack. The ethics review must span the entire hardware-software ecosystem — from the environmental and labour conditions of hardware extraction to training-data sourcing decisions — not only the model's behavioural outputs.

Data Sovereignty. Indigenous communities must control data generated from and about them. This principle intersects with the Indigenous data sovereignty framing and the CARE Principles for Indigenous Data Governance, while originating from a distinct AI-design-practice register (see Position within the corpus below).

The "relation-oriented AI" framing

The concept that unifies the paper's heterogeneous contributions is what the 2023 Debates essay terms "relation-oriented AI" — a design approach that centres kinship networks, community protocols, and the relationships between human and nonhuman entities as the foundational design register, rather than individual user preferences or optimisation metrics. The Kanaka Maoli framing quoted at the opening of the guidelines — "Aloha is the intelligence with which we meet life" — anchors this concept: aloha, understood as love, affection, compassion, and relational greeting, is named as a form of intelligence rather than a sentiment adjacent to intelligence. The implication is that an AI designed from within Kanaka Maoli epistemology would model aloha-style relational intelligence, not the strip-mining of pattern from decontextualised data that characterises contemporary large-scale model training.

Lewis has described the paper's wider ambition as moving from "impoverished intelligence" — AI systems built on a narrow band of Western-rationalist epistemologies that systematically exclude and misrepresent non-white, non-male, and non-Western peoples — to "abundant intelligences," a design frame that treats the full spectrum of human knowledge traditions as a resource for building AI that contributes to the flourishing of all humans and non-humans.

Abundant Intelligences research program

The position paper's working group evolved into the Abundant Intelligences research program, an Indigenous-led, Indigenous-majority international interdisciplinary initiative co-directed by Lewis and Hēmi Whaanga. In 2023 the program was awarded a New Frontiers in Research Fund (NFRF) Transformation grant of more than CAD $22 million, with a team of 37 co-investigators and collaborators from eight universities and twelve Indigenous community-based organisations in Canada, the United States, and New Zealand, the majority of whom are Indigenous. The scale of the Abundant Intelligences grant is the institutional indicator of the position paper's uptake: the 2019 workshop project became the evidence base for a flagship federal-research-fund investment in Indigenous-led AI design.

Position within the corpus

The IP AI Position Paper sits in a distinct register from The CARE Principles for Indigenous Data Governance, the corpus's other foundational Indigenous-led publication in this space. CARE addresses the governance of data about Indigenous peoples — the political economy of data collection, ownership, control, and reuse — and is grounded in Indigenous data sovereignty as a rights claim under UNDRIP. The IP AI Position Paper addresses AI design from Indigenous knowledge systems — what kind of intelligence AI should be modelled on, what relational and territorial obligations the design process carries, and how Indigenous protocols should structure AI governance from the ground up. The two documents are complementary rather than overlapping: CARE asks "who governs data about us?" while IP AI asks "what should AI be if it is to serve communities like ours?" Both are grounded in the framing that Indigenous data sovereignty names, and both sit in the lineage of Indigenous-led technology practice of which Te Hiku Media's Kaitiakitanga License is the working-code expression.

As a publication type the IP AI Position Paper fills the position paper / working-group report slot — a substantive multi-contributor document rather than a peer-reviewed paper, manifesto, or book — and its 2020 publication date makes it the corpus's earliest AI-design-specific Indigenous-led foundational text. Its combination of design guidelines, art, and technology prototypes in a single document is the most direct precedent for the broader argument that AI ethics must not be severable from the knowledge traditions and cultural forms of the communities whose lives AI shapes.