This post analyses public preferences of the use of GenAI in the policy decision-making process across New Zealand and compares non-New Zealand Māori to other New Zealand’s preferences. The methodology is at the end of the article.

Public preferences of the use of GenAI in the policy decision-making process across New Zealand.

Each column is a ranking from 1 to 20. The first column shows how Māori respondents ordered twenty things they want from generative AI, from most important (1) to least important (20). The second column shows how non-Māori respondents ordered the same twenty things. Comparing the two columns tells us where the groups agree and where they part ways.

Question	Māori	Non-Māori
Personal data privacy and security	1	1
High accuracy	2	3
Global regulatory compliance	3	4
Human oversight	4	2
National regulatory compliance	5	7
Job preservation	6	5
Transparency in use	7	8
Misinformation monitoring	8	6
Māori data sovereignty protection	9	19
Māori expert involvement	10	18
Intellectual property compliance	11	9
Local data training	12	10
Māori commercial integration	13	20
Open-source transparency	14	12
Domestic private ownership	15	16
New Zealand commercial integration	16	11
Owned by an Aotearoa/New Zealand government entity	17	15
Faster decision-making	18	13
Energy efficiency	19	17
Cost efficiency	20	14

Māori and non-Māori want the same things. This is reflective of other studies from the Privacy Commissioner and InternetNZ about New Zealander’s perspectives of Artificial Intelligence.

Key observations

• Both groups put personal data privacy at number one.
• Both put accuracy, human oversight, and compliance with regulations near the top.
• Both care about jobs, transparency, and misinformation.
• Across the top eight items, no attribute moves more than two places between the two groups.

Māori and non-Māori share several core concerns about AI and data, particularly privacy, accuracy and human oversight, although existing NZ privacy research indicates that Māori can experience or express these concerns more strongly than non-Māori.

Three attributes are specifically about Māori:

1. protecting Māori data sovereignty
2. involving Māori experts
3. integrating Māori commercial interests.

These are the only items where the two groups seriously diverge significantly, as would be expected. Māori respondents put these three items at 9th, 10th, and 13th. Non-Māori respondents put the same three items at 19th, 18th, and 20th.

For non-Māori, protecting Māori data sovereignty matters less than the AI being cheap, less than it being fast, and less than it being energy efficient.

The items non-Māori rank much higher than Māori do are cost efficiency, faster decision-making, and commercial opportunity for New Zealand businesses.

How AI is governed in New Zealand.

Māori comprise 17.8% of the population by Māori ethnicity in the 2023 Census (and 19.6% by Māori descent). If the government or industry decides AI policy by asking ‘what do New Zealander’s want?’ through surveys, market research, or public consultation where every voice counts equally, then Māori data sovereignty will land near the bottom of the list every single time, because the roughly 82% of respondents who are non-Māori ranked it near the bottom on average 19th out of 20.

This might change if discussions and awareness about New Zealand sovereignty and security of data become more common.

Priorities

Māori respondents ranked Māori Data Sovereignty 9th, behind privacy, accuracy, and oversight. Someone will inevitably say, “even Māori don’t prioritise this so why should policy?”

First, a ranking exercise forces people to put things in order, which creates artificial trade-offs. Ranking data sovereignty 9th means privacy and accuracy are also essential. A parent asked to rank ‘food, shelter, education, safety’ for their children isn’t saying education is dispensable by putting it fourth. All the top-ranked items can be non-negotiable at once; the survey format just can’t show that.

Second, the categories themselves may be splitting one Māori concern into pieces. When a Māori respondent puts ‘personal data privacy’ at number one, that answer likely carries meanings the survey can’t capture, data as a taonga, information connected to whakapapa, obligations that extend beyond the individual to whānau, hapū, and iwi.

The Western framing of ‘personal privacy’ and the separate item ‘Māori data sovereignty’ may, for Māori respondents, be two labels for overlapping parts of the same thing. If so, the survey is spreading one concern across two items and making each look weaker than it is.

The gap between 9th and 19th could reflect a huge difference in how many people endorsed the item, not how far apart the items are. The underlying percentages would tell us which it is, and they’re worth getting hold of.

Conclusion

Māori and non-Māori are nearly indistinguishable in wanting the same foundations first from Gen AI: privacy, accuracy, human oversight, and adherence to the law.

The divergence appears only on the three explicitly Māori attributes: data sovereignty, expert involvement, and commercial integration, which non-Māori respondents rank at the very bottom.

This creates a structural problem for how AI policy is decided. If governance is set by aggregating what New Zealander’s want, weighting every voice equally through surveys and consultation, Māori data sovereignty will land near the bottom every time, not because it lacks merit, but because the roughly 82% of respondents who are non-Māori rank it near the bottom on average.

Majoritarian aggregation doesn’t neutrally reflect preferences; it structurally deprioritises the interests of a minority that makes up 17.8% of the population.

The objection that ‘even Māori ranked it only 9^th’ is misleading. A forced ranking manufactures trade-offs between items that may all be non-negotiable, and the survey’s Western categories likely fracture a single Māori concern across separate labels, so that ‘personal data privacy,’ placed first, already carries meanings of taonga, whakapapa, and collective obligation that the instrument records elsewhere as ‘Māori data sovereignty.’

Policy makers must consider Te Tiriti obligations.

 

Methodology

The statistics are a modified set of data from an unpublished document “Public preferences of generative artificial intelligence in policy decision-making in Aotearoa New Zealand. Yuelu Xu, Maksym Polyakov, David Worden, Karaitiana Taiuru, Dan Richards. The methods and data sections remain in the original work. For this article, Xu Yuelu was asked to extract the Māori data to compare to non-Māori data. This request is shown in the comparison table and is not modified.

All of the commentary is the authors own analysis and conclusions.