Earlier this week, the New Zealand Government published its Algorithm Charter for Aotearoa New Zealand to guide the use of algorithms by its agencies.
Building on previous artificial intelligence (AI) research by the New Zealand Law Society, the Charter emphasizes that more complex algorithms can be used to support human decision-making, helping the Government to better understand New Zealand and New Zealanders. It goes on to recognize that a principled approach is needed to realize this potential and mitigate the risk of unintended consequences, and as such, government agencies that sign onto the Charter commit to following its principles when using algorithms to help serve the people of New Zealand.
The Charter takes a leading step forward in its human-centred approach: it embeds Te Ao Māori perspectives and contemplates active participation by impacted communities. We welcome the Charter’s upfront commitment to transparency and human oversight, and its proportionate, risk-based approach to the deployment and use of algorithms by government agencies.
With the Charter, the New Zealand Government is setting a strong foundation for guiding its agencies on how to implement algorithms in a manner that warrants trust, considering the key elements of Transparency, Partnership, People, Data, Privacy/Ethics/Human Rights as well as Human Oversight.
Additional considerations
As part of the New Zealand Government’s policy-making process, Microsoft provided official comments on an earlier draft of the Charter. We believe that the principles outlined in the finalized Charter are of fundamental importance, and that going forward, it will be equally important to work with signatories to help operationalize them.
Take the principle of transparency as an example. Different stakeholders have different transparency needs that will rarely be met by a single approach. Selecting a transparency approach means focusing on the people at the centre of an algorithm’s development and use, and the goals that transparency might help them achieve (e.g., improved robustness, better usability, or increased trust). For example, there are scenarios where transparency is needed to help decision-makers like doctors or judges decide when and how much to trust AI systems. That differs from the transparency needed to help uncover potential sources of bias or unfairness, or how to convey how systems behave in different conditions.
In our comments on the earlier draft, we brought to the New Zealand Government’s attention that there has been progress in developing tools to understand and manage some of the socio-technical challenges associated with responsible AI, including detecting uses that might result in unfairness, traceability through model management, AI security, and improving intelligibility. We recommend agencies leverage these efforts in their forthcoming work to operationalize the Charter. Our Responsible AI Resource Center may be helpful in this endeavour.
As a next step, we’d encourage the New Zealand Government to consider developing practical implementation guidelines, including sharing examples of government projects that have piloted the principles of the Charter. We recommend providing government agencies that are expected to apply the Charter with real-life examples that illustrate the technical, organizational and policy safeguards could help them deliver on the Charter’s objectives. Microsoft welcomes the opportunity to provide some of our learnings from operationalizing our AI ethics principles as the New Zealand Government may look to best practices observed in other countries or organizations.
It is also valuable for agencies to become familiar with the principles and embed them into their work – especially as algorithms are used by agencies to provide useful services to New Zealand citizens. The Charter will be even more meaningful as additional agencies sign on.
Peer review of the effect of algorithms can be helpful. We believe that such a review is most effective when it considers the operation of the algorithm in the real-world. Real-world scenarios will provide the best gage of potential unintended consequences and allow for robust consideration of what steps are needed in response.
International context
As the New Zealand Government has been taking this important step forward with its Algorithm Charter, the European Union has been spearheading the development of a comprehensive policy approach to AI. The framework designed by the European Commission aims to allow for industry to use AI responsibly to innovate and to help people prepare for using technology to augment their abilities as more organizations and businesses embrace digital transformation. Microsoft made a number of suggestions on how the policy can be strengthened and expects the European Commission to publish a regulatory proposal in the coming months. As New Zealand may consider its own national AI strategy, we encourage the Government to stay informed about the process in Brussels.
In the United States, Washington State recently enacted the first law in the nation to establish specific guardrails on the use of facial recognition services by state and local government agencies. This law demonstrates important progress on determining whether and how facial recognition services should be used. We called for regulation of this technology in 2018, and we’re encouraged that the law in Washington state is a model that legislators can build on.
At Microsoft, we continue to use our facial recognition principles to guide how we develop and deploy our technology. The principles are centred on fairness, transparency, accountability, non-discrimination, notice and consent, and lawful surveillance. We also introduced a policy to not sell facial recognition technology to U.S. police departments until there is a strong national law, grounded in human rights. In the coming year, we will continue to engage on this issue with legislators, industry, and civil society groups, among others, in countries around the world and we will advocate for regulation in both government and commercial contexts to prevent a commercial race to the bottom where people are left without protection under the law.
Finally, the Asia Pacific Economic Cooperation’s (APEC) Business Advisory Council (ABAC) has also raised attention to the importance of regional collaboration to create an environment for the successful development and ethical uptake of AI. As New Zealand assumes the role of chair and host for APEC in January 2021, the Government will be well-placed to work with other economies in the region as well as ABAC to foster dialogue in this domain – in particular, by exporting New Zealand’s strong ethical standards.
In this context, the New Zealand Government’s approach of starting with self-regulation of its own agencies via frameworks such as the Algorithm Charter is quite unique. The policy has the potential to showcase to other organisations in Aotearoa what the North Star in responsible algorithms deployment should be.
We look forward to assisting agencies implement the Charter, operationalize its principles, and iterate on future versions (for instance, to incorporate learnings and gradually expand its scope). We’re committed to the human-centred development of AI that helps us better serve New Zealand and New Zealanders.
Maciej Surowiec
Government Affairs Lead – New Zealand
Corporate, External & Legal Affairs (CELA)
To learn more about Microsoft’s approach to Responsible AI, click here.