MD_OPS 3: How do you use or share Māori data back with Māori? - Māori Data Lens

MD_OPS 3: How do you use or share Māori data back with Māori?

Organisations use data in many different ways to design and deliver products and services. They can use it to gain insight and understanding of their organisation, their industry, and the wider world around them. Organisations should assess any legal and ethical implications when making decisions relating to how data will be used and shared.

  • MD_OPS03-BP01: Collecting and separating Māori data appropriately. Organisations can consider how to collect and separate data along different dimensions such as iwi and hapū or a Māori organisation. This may make data more relevant or useful to different groups or communities. This needs to be considered when designing your initial data collection plan to verify that the right data is collected and organised early. For example, a government agency may want to report on specific agency outcomes for Māori populations. If they captured an individual's hapū affiliations, they could report information at a hapū level. However, if they only captured an individual's record, then they can only report this at an individual level.

  • MD_OPS03-BP02: Consider how your organisation could share Māori data back to Māori. Data that your organisation holds could be used to better understand Māori communities and realise individual and collective benefits. Consider how you could identify useful data, and design ways to make data more accessible. Open data initiatives are one approach, but also remember the importance of complying with the Privacy Act 2020.

  • MD_OPS03-BP03: Use tools to effectively and securely share data where there is a specific and appropriate purpose. There are many approaches to sharing data. Consider tools to share data both publicly and privately, which fits the purpose for why that data is being shared and how it needs to be used. Share this data in culturally-appropriate ways, and avoid misappropriating that data or trying to create explanation around that data that isn't culturally-sensitive or informed. Seek advice from your Māori advisers if you are unsure how to do this. This includes open data portals or exchanges such as the AWS Open Data Registry, AWS Data Exchange, or private data portals hosted by your organisation. Consider solutions that allow data to be shared using different formats such as flat files, application programming interfaces (APIs), or interactive dashboards and visualisations to meet the needs of data consumers.

  • MD_OPS03-BP04: Share data in ways that can be easily used by your various stakeholders. Understand your Māori stakeholders, their interests in accessing and using the data you are sharing back with them, and how they use the data. This should drive your choice of format for that data. For example, graphs or visualisations published on your website can make data easy to find and understand, while data files such as comma-separated values files (CSV) or parquet are better for data analytics users. An API supports application-to-application integration. For example, a government agency may provide interactive graphs on their website so that anyone with a web browser can see graphs relating to key agency objectives. They may provide the data that sits behind the graphs in a CSV format so that people can download it and load it into a spreadsheet tool or analytics programme to build their own graphs or perform additional queries. They may also provide an open API so members of the public can retrieve the data and load into their own databases or analytics systems.

  • MD_OPS03-BP05: Consider how your organisation could use federated data access methods. Organisations often need to access data from other organisations to support a business process. Federated data approaches can allow organisations to access data from other organisations without the need to replicate or copy the data into your own systems. Federated data access models require appropriate mechanisms for making your organisations data discoverable and for securely sharing data.

  • MD_OPS03-BP06: Define and implement appropriate authorisation mechanisms. Design data access mechanisms that support both internal access and access for external third parties (for example, through federated data access or data sharing mechanism). Consider authorisation mechanisms including role-based access control (RBAC), attribute-based access control (ABAC), or policy-based access control (PBAC). The authorisation mechanism should provide ways to manage, grant, and revoke access and provide visibility into data access through auditing and logging. Establish appropriate governance processes to effectively manage the process for requesting, granting, and revoking access to data by verified, trusted, and approved external third parties.

  • MD_OPS03-BP07: Incorporate responsible and ethical use of machine learning (ML) and artificial intelligence (AI) as a core part of your governance framework and development lifecycle. ML and AI have transformational potential. It is already widely used for tasks such as transcription, translation, fraud detection, information security, search, and recommendation engines. At Amazon, we believe the design, development, and deployment of AI must respect the rule of law, human rights, and values of equity, privacy, and fairness. We are committed to developing fair and accurate AI services and providing customers with the tools and guidance needed to build applications responsibly. Developers and deployers of AI systems should ensure such systems are built based on principles of safety and responsibility by design. AWS builds AI with responsibility in mind at each stage of our comprehensive development process. Throughout design, development, deployment, and operations, we consider a range of factors including accuracy, fairness, appropriate usage, toxicity, security, safety, and privacy. Ask your Māori customers what particular ethical questions and principles they may want you to adopt as a part of your governance framework and development lifecycle. Regularly revisit and refine your AI ethics practices through ongoing dialogue and partnerships with Māori communities.

  • MD_OPS03-BP08: Leverage vendor tools to provide AI transparency. For example, AWS AI Service Cards deliver a form of transparency documentation that provide customers with a single place to find information on the intended use cases and limitations, responsible AI design choices, and deployment and performance optimisation best practices for our AI services. Amazon SageMaker AI Clarify detects and measures potential bias using a variety of metrics so developers can address potential bias and explain model predictions. Amazon's Responsible Use of Machine Learning Guide highlights key best practices and tooling that AI developers and deployers can use to mitigate risks across the lifecycle of an AI system.

  • Other considerations:

    • Discuss AI with stakeholders. Artificial intelligence including generative AI and machine learning is a rapidly evolving area. Discuss the benefits and risks with your stakeholders. Your stakeholders may be internal to your organisation, external customers, or members of the public. Cross-functional expertise from technologists, ethicists, lawyers, domain experts, and external resources provides a holistic understanding and consideration of ethical, legal, and domain-specific factors. Customers should engage with Māori stakeholders early and continuously to understand their perspectives, concerns, and preferences regarding the use of certain data in AI/ML deployment and development. Employ Māori expertise in the design, development, and testing phases to verify cultural competence and alignment with tikanga.

    • Consider potential inaccuracies. Customers should consider potential inaccuracies in ML system results (especially concerning te reo Māori and Māori cultural concepts) and prepare a plan to address them, such as narrowing scope, introducing human oversight, or altering dependencies on the AI system. Customers should also consider incorporating appropriate testing of model outputs into the AI application creation process when model inputs or outputs are dealing with te reo Māori or Māori cultural concepts. Evaluation of outputs should be made by appropriately skilled people. To assess if an AI system operates as intended, it is important to use accurate and representative training data. AWS encourages specific policies and provides safeguards such as Guardrails for Amazon Bedrock to block harmful user inputs.

    • Use case evaluation and testing. Testing should include not just the AI system itself but also the overall process it is a part of, including decisions or actions that might be taken based on system output. Customers should consider evaluating models thoroughly on safety characteristics such as prompt stereotyping (like encoded biases for gender or socioeconomic status), factual knowledge, and toxicity. FMEval, an open source library is available in Amazon SageMaker AI for developing these insights. Model evaluation is also available in Amazon Bedrock for large language models (LLMs). Model outputs can be evaluated by a pool of human evaluators, or through an automated process. Customers can test performance through techniques like red teaming and reinforcement learning from human feedback (RLHF). Customers should also consider continually evaluating performance and responsibility metrics before deployment and use tools like SageMaker AI Model Monitor to detect data drift and prompt retraining if needed.

    • Continuous improvement and validation. Monitoring for potential bias and accuracy, and for models performing as expected across different segments, is an important part of this process.

    • Ongoing education. AI is a constantly-evolving landscape, and new techniques, technologies, laws, and social norms continue to be developed and refined over time. It is essential that all parties involved with building and using AI systems stay educated on these issues and account for them in the design, deployment, and operation of their systems. Successful AI adoption requires significant cultural and organisational changes, including defining the roles and responsibilities required for accountability, and customers should verify that they are allowing Māori to make informed decisions about participation.

For further reading, refer to the AWS Machine Learning lens.

AWS continues to update this information and share additional guidance to customers on the use of AI/ML. Please reach out to the team at AWS for further updates.