How New Zealand is creating open algorithms

By Shirley Tay

Craig Jones, Deputy Chief Executive of Stats NZ, discusses the government’s commitment to transparency.

Governments have collected vast amounts of data in the fight against Covid-19. Contact tracing tools and self-reporting apps, for instance, require information on personal meetings - underscoring the need for privacy controls.

New Zealand’s government is leading an effort to be transparent in its use of data. Stats NZ, the country’s data agency, conducted an assessment since 2018 to better understand how algorithms are being used by the government.

“At the heart of it, algorithms are not really about data - they're about people. So how do you engage people in its development?” Craig Jones, Deputy Chief Executive at Stats NZ, discusses how his team is improving transparency in the government’s use of algorithms.

Enforcing transparency

Stats NZ’s assessment found that algorithms “by and large, did really useful things for making government more efficient and effective,” Jones says.

But there wasn’t a “great deal of transparency” on how data was being used. Without that accountability, it was difficult for the public to have confidence in how the government was using data, he adds.

Stats NZ developed a set of standards last July to guide how agencies should use algorithms. 26 departments - such as the Ministry of Justice and the Department of Corrections - have committed to the voluntary charter as of today.

The charter requires agencies to be publicly transparent on what algorithms are driving decision-making, how data is used, and to explain it in ‘plain English’. Agencies must also ensure data quality issues are not leading to biases, and put checks and balances in place so they “don’t just create an algorithm and let it run”.

New Zealand’s commitment towards open algorithms is also part of the Open Government Partnership, a multilateral initiative to expand citizen participation in government.

The New Zealand Police is one of the signatories of the charter. It had come under controversy last year for trialling facial recognition tech without announcing it or consulting the Privacy Commissioner, Radio New Zealand reported.

“People have every right to be worried about how images of their face are being taken and used … I think it's important that there's transparency around that,” Jones says.

After signing and committing to the charter, the Police has published their own “stocktake of how they're using these kinds of technologies”. This helps citizens understand what they’re doing and why they’re doing it, he adds.

Stats NZ plans to review the charter’s progress in the second half of 2021, Jones says. “Until then, we’re really taking a more facilitation, coordination, guiding kind of role.”

Partnering with the Māori

As part of the charter, agencies must also ensure there is partnership with New Zealand’s indigenous people, the Māori. “Often, the decisions made have a disproportionate impact on Māori people. They're under-represented in a lot of the statistics that we publish,” he adds.

Stats NZ helps agencies understand how to involve the Māori in the decision-making process. The agency is working with its training partners to co-design a “governance model that can be widely shared across the system”, Jones says.

Ultimately, “it’s really up to the individual agencies, through the relationships that they hold, to work out how to engage people,” he says. “They're in the best place to know how to do that and be accountable for doing them.”

Finding the right balance

In the process, Stats NZ faced a challenge in defining what an algorithm is. “Every Excel spreadsheet has algorithms sitting behind them,” Jones says.

“If you go too wide and require agencies to document every single algorithm they use, you'd be bogged down by bureaucracy - writing documents for a website rather than actually doing the work that the public expects you to.”

For the charter, Stats NZ decided to use a risk-impact assessment. If the algorithm involves something risky and of public interest, the agency should be transparent about it, he adds.

Finding the middle ground has been another challenge for his team. Some still want “much greater regulation around the use of algorithms”, while for others, “we’re getting too far involved”.

“I think we've done a reasonably good job of that for now,” Jones says. He acknowledges that if things or contexts change, his team will have to rethink that delicate balance.

His team also discovered that there’s no “one size fits all” approach, he says. What works for the Inland Revenue Department will be very different from what works for the Ministry of Social Development. “We need to think quite laterally about how we support agencies, rather than getting in their way.”

Look to existing guidelines

For governments looking to embark on a similar commitment, Jones says it’s important to keep in mind that they’re “not starting from zero”. Organisations can build on what’s already done before - whether it’s the AI ethics guidelines global inventory or Stats NZ’s work.

But what works in one jurisdiction won't necessarily work in another, and governments need to be willing to “iterate, test, learn and adapt as they go”.

New Zealand is “only scratching the surface” of the potential of algorithms in decision making, Jones believes. As data powers more policy decisions, the country’s commitment to transparency will go a long way.