Skip to content

Developing the Charter – Report: An Algorithm Charter for Aotearoa New Zealand

This briefing updated you on the submissions received on the draft Algorithm Charter and sought your agreement to discuss a revised Charter at the Digital and Data Ministers meeting on 4 March 2020.

Summary of Submissions
Report: An Algorithm charter for Aotearoa New Zealand [PDF 468 KB]

Contents

Recommended actions

The following are the recommendations that were presented:

  1. Note that we have completed an analysis of the submissions to the 2019 public consultation on the draft Algorithm Charter (the Charter).
  2. Note that the public submissions show a strong level of support for the introduction of a charter and many submitters provided feedback about the Charter’s scope.
  3. Agree to publish the document ‘Algorithm Charter: Summary of Submissions’ (attached as Annex Two).
  4. Note that, based on feedback received from public submissions, we have revised the Charter text to clarify scope, improve alignment to the Treaty of Waitangi, make clearer links to the Principles for the Safe and Effective use of Data and Analytics and strengthen consultation expectations.
  5. Note that, based on feedback received from public submissions and previous feedback from government agencies, we have amended the scope of the Charter to initially apply to all new algorithms, and to apply to all existing algorithms after a period of 5 years.
  6. Agree to circulate this paper to Digital and Data Ministers ahead of their next meeting on Wednesday 4 March 2020.
  7. Agree to seek the endorsement of Digital and Data Ministers to the revised Charter.
  8. Note that, if Ministers endorse the revised Charter, we propose to discuss the initial signatories at the next meeting of the Digital Government Leadership Group Chief Executives on 9 March 2020.

Background

In October 2018 the Government Chief Data Steward and Government Chief Digital Officer published an assessment of the use of algorithms in 14 government agencies. The Algorithm Assessment Report made several recommendations to help New Zealanders be informed and have confidence in how the government uses algorithms to inform decisions about individuals

In June 2019 we advised you of options to respond to these recommendations. We recommended a combination of a voluntary algorithm charter and workforce capacity initiative. This recommendation was subsequently endorsed by the group of Digital and Data Ministers, and officials were invited to report back to the next meeting of the group with a plan for delivery.

In September 2019, following consultation with government agencies, we outlined a proposal for a voluntary algorithm charter, based on the model used by the government accessibility charter, that would commit government agencies to improve transparency and accountability in their use of algorithms in five years. We proposed a period of public consultation prior to finalising the text of the draft Algorithm Charter (the Charter).

This consultation also delivers on our Open Government Partnership commitment to consult with civil society on our efforts to increase transparency and accountability of how the government uses algorithms.

This recommendation was subsequently endorsed by the Digital and Data Ministers, and the Chief Executives who form the Digital Government Leadership Group. We were invited to report back to the next meeting of the Digital and Data Ministers, following public consultation, and seek agreement for the implementation of the Charter in early 2020.

On 17 October 2019, you launched the Charter for public consultation, with submissions to close at the end of 2019.

Other work to improve government transparency and accountability is ongoing

The Charter forms a part of our wider work programme to improve transparency and accountability around government data use, including:

  • The ongoing operation of the Data Ethics Advisory Group to provide government agencies with expert advice on new and emerging uses of data.
  • An initiative to improve data ethics in the government analytics workforce.
  • Consideration of whether change is needed to the current system settings of government use of data to build public trust.

A voluntary charter provides an important level of consistency

One of the key findings of the Assessment was that, while government agencies are applying a range of safeguards and assurance processes in relation to their use of algorithms, there’s a lack of consistency across government.

The Charter addresses this lack of consistency by providing a clear and articulated expectation of the actions that agencies will take in relation to algorithm transparency and accountability, given effect by a written commitment from Chief Executives and key managers responsible for algorithm development.

During our previous consultation with government agencies, some felt that an incremental approach posed by a voluntary charter was ‘too little’ while others felt that it was too much and risked ‘stifling innovation’.

As a result of the feedback we received, we revised the Charter to be clearer around its scope, implementation and accountability, as well as clarifying where it was positioned amongst other related tools and initiatives such as the Data Protection and Use Policy (published by the Social Investment Agency in December 2019)

On the basis of subsequent discussion across government, it is clear that some government agencies and sectors are well served with guidance, oversight and tools around the use of data and algorithms, while others have no such support .

We believe a voluntary charter strikes the right balance between providing clear, consistent expectations for those agencies who seek them, without being overly prescriptive to agencies who already have well developed approaches to managing algorithm transparency and accountability.

There’s strong public support for a charter on algorithms

We received 34 written submissions on the Charter as part of the public consultation process. Submissions were received from a range of individuals, organisations, and government agencies. The ‘Summary of Submissions: Draft Algorithm Charter’ report is appended to this report as Annex Two.

Most submitters supported the introduction of a charter and provided feedback about defining the Charter’s scope. The key themes from the public submissions are:

  • That the purpose and scope of the Charter should be broadened to include all algorithms, not just operational algorithms.
  • That government algorithm use should reflect the principles of the Treaty of Waitangi and the Charter should specifically consider the protection of Māori interests.
  • General support for monitoring Charter implementation, including several proposals for the establishment of an independent monitoring body.
  • The need for regular review of bias and public reporting of implementation.
  • The need to develop specific guidance to accompany the Charter and aid implementation.
  • That human judgement and evaluation are part of the decision-making process based on data, and this should be addressed in the Charter.

Public submissions provided valuable insight on how this topic is perceived outside of government, and provided many positive suggestions for improvement. In particular, we believe that the scope of the Charter, the way in which Te Ao Māori is described and arrangements for implementation and accountability need further consideration.

Many submissions disagreed with limiting the Charter to operational algorithms

The feedback we received suggests that the public see no reason for the scope of the Charter to be limited to only operational algorithms, however, there was a recognition in some submissions that extending the scope more broadly would likely involve additional cost and complexity for government agencies.

The original rationale for a limited scope was to align the Charter with the Assessment and focus the efforts of government agencies on those algorithms that have the most significant impacts on the lives of people.

Based on the feedback from some large data agencies we understand there is a clear preference to limit the scope of the Charter, due to the resource implications of retroactively reviewing current algorithm settings. They argue that additional processes required to give effect to the Charter may stifle innovation. It is possible, therefore, that extending the scope could reduce the number of agencies willing to sign the Charter.

Based on feedback from government agencies, and public submissions, we’ve identified several possible options for the scope of the final Charter:

  • Operational algorithms. This was proposed for the reasons outlined in the paragraphs above, but many public submissions were critical of this narrow scope.
  • New algorithms. This would allow for a more gradual adoption of the Charter by agencies, reducing the resourcing burden, and allow the Charter to apply to all algorithms.
  • All algorithms. This option was supported by many public submissions, suggesting it would have the greatest impact to increase public trust, but it may have a negative impact on the number of agencies willing to sign as outlined above.

On balance, we recommend that the Charter have a staged implementation. Initially the Charter would apply to:

  • All new algorithms (those not yet in service), including both operational and others, from the date of signature.
  • After a 5-year period the Charter would also apply to all current algorithms used by an agency (including both operational and others).  

This removes the distinction between operational and other algorithms proposed by the draft Charter, in line with the submissions from the public, while also responding to feedback from agencies about the challenge associated with retrospective review.

The current wording around Te Ao Māori needs to be improved

Submitters agreed that it was important to recognise the Treaty of Waitangi. Half of the public submissions supported the concept of embedding Te Ao Māori into algorithm development or procurement, but thought that the Charter could go further and be clearer about what this means.

Several submissions suggested that the Crown’s Treaty of Waitangi obligations be made explicit.  We have consulted with the Office for Māori Crown Relations - Te Arawhiti and revised wording to ensure we align with other references to the Treaty used within the Crown. 

Some submissions also suggested that the Charter needs to reference Māori Data Sovereignty. This is a complex topic, and we are currently working in partnership with the Data Iwi Leaders to consider system-level governance arrangements. We have added an introduction to the Charter to clearly describe that this is not within scope. 

Implementation and accountability were significant issues for the public

A number of submissions raised concerns about how accountability would be enforced once the Charter is signed. Some submissions suggested the establishment of an organisation that could provide monitoring, oversight on implementation by government agencies and advice and support to help adoption.

The system settings for the devolved model of public service in New Zealand mean that overall accountability for agency decision making across a wide range of functions, including algorithms, sits with individual Chief Executives, who are in turn accountable to the State Services Commissioner.

Checks and balances exist in the form of external regulators who can investigate and respond to complaints about the conduct of government officials and agencies, through the Ombudsman and in relation to the privacy of individuals, through the Privacy Commissioner.

The Assessment showed that operational algorithms are currently employed to automate existing services or create new services within the existing portfolio responsibilities of government agencies. In these cases, the agency responsible has an existing level of accountability for that service and an established process to resolve complaints or requests for information. These algorithms would also be subject to the checks and balances referenced previously.  

For this reason, the Charter serves to make explicit and consistent the commitment of government agencies to a level of good practice, through the signature of the accountable person (the Chief Executive), rather than establish a new process or organisation. However, we have revised the Charter to require signatories to designate a clear point of contact for algorithm issues and a channel for challenging or appealing decisions informed by algorithms.

We have revised the Charter to reflect the feedback we received

In addition to the changes referenced above, we’ve significantly revised the text of the Charter on the basis of the feedback we’re received. The changes include:

  • clearer alignment with the Principles for the Safe and Effective Use of Data and Analytics
  • strengthening the commitment for agencies to actively consult with people, communities and groups interested in, or impacted by, algorithms
  • more accessible plain English text. We will also publish a te reo Māori translation of the final text.

A revised copy of the Charter text is appended as Annex One. We believe that the revised Charter is a significant improvement on the previous draft.

We seek your agreement to circulate this paper, and discuss the revised Charter, at the next meeting of the Digital and Data Ministers on Wednesday 4 March 2020. If Ministers endorse the revised Charter, we propose to discuss the initial signatories at the next meeting of the Digital Government Leadership Group Chief Executives on 9 March 2020.

Agencies will require support to implement the Charter

Government agencies will require support and guidance to implement the Charter. We already have some elements of this support in place, through the operation of the Data Ethics Advisory Group, who can provide agencies with advice on ethical considerations of the new algorithms they develop.

We anticipate developing and publishing a further suite of advice in partnership with agencies, as new algorithms are developed in alignment with the Charter commitments. This work will be enabled by the additional funding to the Statistics Multi-Category Appropriation made through Budget 2019.

A key aspect of public feedback not currently addressed by the revised Charter is the need to have a central point that provides monitoring and reporting of algorithms. By adopting a staged approach, with an initial focus on new algorithms, it will be possible for the Government Chief Data Steward to maintain and publish a quarterly list of new algorithms supplied by signatories. This is well aligned with our current work on lifting the transparency and accountability of government data use.

We will provide you with further advice on options to the above proposed function for all government algorithms employed by signatories ahead of the 5-year Charter deadline. This advice will form part of the next paper considering possible system changes to build public trust in government use of data. 

Next steps

We seek your agreement to publish the Algorithm Charter: Summary of Submissions document appended to this paper as Annex Two.

If you agree with the revised Charter, we will circulate this paper to Digital and Data Ministers ahead of their next meeting on Wednesday 4 March 2020.

At that meeting we recommend that you seek the endorsement of Digital and Data Ministers to the revised Charter.

If Ministers endorse the revised Charter, we propose to discuss the initial signatories at the next meeting of the Digital Government Leadership Group Chief Executives on Monday 9 March 2020 and will work with your office to coordinate a public announcement.

Annex One

Introduction

This Charter demonstrates a commitment to ensuring New Zealanders have confidence in how government agencies use algorithms. This Charter is one of many ways that government is demonstrating transparency and accountability in the use of data. However, it cannot fully address important considerations, such as Artificial Intelligence and Māori Data Sovereignty as these are complex and require more focussed solutions. 

Commitment

Our organisation understands that decisions made using algorithms impact people in New Zealand.

We commit to the actions outlined in this charter in developing all new algorithms. 

Over the next five years, we commit to aligning existing algorithms with the actions outlined in this Charter and the Principles for the Safe and Effective use of Data and Analytics.

Committing to using algorithms fairly and transparently means we will: 

Maintain transparency by:

  • Clearly explaining how decisions are informed by algorithms, this may include:
    • plain English documentation of the algorithm
    • making information about the data and processes available (in some cases, such as algorithms being used for national security, there may be exceptions)
    • publishing information about how data are collected, secured and stored.
  • Deliver clear public benefit through Treaty commitments by:
    • embedding a Te Ao Māori perspective in the development and use of algorithms consistent with the principles of the Treaty of Waitangi.
  • Focus on people by:
    • identifying and actively engage with people, communities and groups who have an interest in algorithms, and consult with those impacted by their use.
  • Make sure data is fit for purpose by:
    • understanding its limitations
    • identifying and managing bias.
  • Ensure that privacy, ethics and human rights are safeguarded by:
    • regularly peer reviewing of algorithms to assess for unintended consequences and act on this information.
  • Retain human oversight by:
    • nominating a point of contact for public inquiries about algorithms
    • providing a channel for challenging or appealing of decisions informed by algorithms
    • clearly explaining the role of humans in decisions informed by algorithms. 

Annex Two – Algorithm Charter: Summary of Submissions

Algorithm charter: Summary of submissions

Contact us

If you’d like more information, have a question, or want to provide feedback, email datalead@stats.govt.nz.

Content last reviewed 20 November 2020.

Top