The Institute of Environmental Science and Research Ltd (ESR) has developed a Responsible AI Framework to guide the safe and reliable development and use of AI and automated workflows. This Framework supports ESR to align with and implement the principles of the Algorithm Charter for New Zealand. The Charter was released by the Government Chief Data Steward (GCDS) In 2020, providing a cross-government commitment to ensure the safe and transparent use of algorithms.
Government uses data to inform decision-making and deliver services to New Zealanders every day. Algorithms play an important part in this, allowing us to break down information from large or complex data sets and reveal insights that could not easily be revealed by human analysis alone. As the Crown Research Institute specialising in science for communities, ESR is exploring how responsible use of AI in combination with its expertise in health, forensic, food, water and radiation sciences can support communities to thrive and prosper.
An independent review of the Algorithm Charter, published last year, provided many insights, not least being that the Charter, across its 27 signatories and beyond, has been influential.
In the case of ESR, its Responsible AI Framework guides the organisation’s approach to developing algorithms that support and enhance the delivery of the science expertise it provides. Current use cases include focusing on AI opportunities to improve the throughput of cases within their laboratory (for example, automatically identify sperm cells in sexual assault cases), and to improve the accessibility of science for frontline responders (such as automated real-time screening of suspected drug samples for Police).
The Responsible AI Framework supports ESR to apply the Algorithm Charter in its own context by supporting projects to consider the risks and benefits of this technology. It provides guidance on how ESR can make the most of emergent technology to increase its impact, while maintaining the confidence and trust of stakeholders.
The Responsible AI Framework includes a Guidelines document that describes the purpose and process, a Questionnaire defining important considerations across a project’s lifecycle, and a Register to document AI activities underway and implemented. By using these resources, ESR staff and their projects are guided through an assessment and approval process to consider potential risks and document mitigations through the five key stages of an AI project: design, development, piloting, implementation, and monitoring.
At the core of the Responsible AI Framework is the recognition that AI can be a powerful tool for ESR’s role in delivering scientific expertise across health, environment and justice for the communities it serves, but only if it is used safely and appropriately.
This is a great example of the direction set by the Algorithm Charter, and how government departments, agencies and crown entities are applying its principles in their own contexts and environments for the benefit of New Zealanders.
For more information on ESR’s Responsible AI Framework, please contact Dion Sheppard at dion.sheppard@esr.cri.nz.