Module 5 of 6
Hoahoa tika
Designing ethical systems — what does a decolonised AI architecture look like?
fish scales — abundance, interconnection, ethical design in layers
This module carries unaunahi — fish scales, representing abundance and connection to the moana. Ethical design, like unaunahi, works in layers: each element interlocking with the next, no part working alone. Every layer of a system carries responsibility for the whole.
He kupu whakataki — Introduction
The previous modules have diagnosed the problem. This one turns to practice: what would it actually mean to design AI systems in accordance with tikanga Māori and Indigenous data sovereignty principles? What are the design requirements, and what would the architecture look like?
Hoahoa tika — ethical design — is not a checklist. It is a practice of putting values into structure: of building systems whose behaviour reflects the relationships and obligations they are meant to serve.
Design principles for Māori-centred AI
- Consent architecture — data should not be collected, stored, or used without clear, ongoing, and revocable consent from the communities it came from. Consent must be collective as well as individual.
- Tapu-aware access controls — systems must distinguish between different kinds of knowledge with different access requirements. Not all Māori data is the same: some is publicly shareable, some restricted, some sacred.
- Benefit-sharing by design — if AI systems generate commercial value from Māori data or knowledge, that value must flow back to Māori communities by design, not by exception.
- Accountability to community, not just regulators — audit mechanisms must be accessible to iwi and hapū, not only to government agencies and corporate compliance teams.
- Te reo Māori as a first-class language — AI systems operating in Aotearoa must treat te reo Māori as a primary language with appropriate cultural context, not as a low-resource curiosity.
Case study: a hypothetical Māori health AI
Consider a health AI designed to assist with Māori health outcomes. What would hoahoa tika require?
- Training data collected under iwi data sovereignty agreements, with ownership retained by the contributing communities.
- The model auditable by Māori health researchers, not only by the company that built it.
- Outputs tested against Māori health values, not only biomedical accuracy metrics.
- Revenue including a benefit-sharing agreement with iwi data partners.
- A tikanga Māori ethics committee — not just a standard IRB — with oversight of the system's development and deployment.
The limits of "ethical AI" discourse
Much of what passes for "ethical AI" in mainstream tech discourse is cosmetic. Bias audits, diversity statements, and advisory boards do not change the fundamental ownership structure of AI systems. Hoahoa tika is not satisfied by ethics washing — it requires structural change in who owns, governs, and benefits from AI.
Key concepts
- Hoahoa tika Ethical design; designing systems that embody tikanga values
- Consent architecture Building meaningful, revocable, collective consent into data systems by design
- Ethics washing Superficial ethics frameworks used to legitimise extractive systems
- Benefit-sharing Mechanisms ensuring value generated from community data flows back to the community
Pātai — Discussion questions
- Design a data consent protocol for an AI system that uses oral histories from kaumātua. What would genuine informed consent look like?
- How does the concept of tapu challenge the design assumptions of standard cloud computing architecture?
- What is the difference between 'ethical AI' as practised by major tech companies and hoahoa tika as described in this module?
- Who should have veto power over the deployment of an AI system that affects a Māori community? Justify your answer.