We are pleased to share the Oxford Insights Trustworthy AI Assessment tool. This self assessment tool is designed to help policymakers in government understand how prepared their government is to use AI in a trustworthy way in the public sector.
Why did we create this tool?
Governments are increasingly using AI in their public services. AI systems are being used to support public services provision, to reshape internal government operations, and to support public priorities outside of government by enabling more efficient and effective use of resources, such as in healthcare, business, and the environment.
Governments need to establish practices and put in place safeguards that ensure AI systems work for and protect various groups, including:
end-users of public services;
civil and public servants;
and the general public who can be impacted by government projects.
This is a new challenge for public sector teams because AI systems present unique risks to manage. How can governments be transparent with users about how high-stakes decisions are made, such as about eligibility for social benefits? How can they evaluate the security and robustness of procured AI systems, especially when these systems are not interpretable or procurement officers do not have the tools or access needed to assess them?
Internationally, there has already been important work on what high-level principles governments should uphold when using AI systems, including from OECD, UNESCO, and the G7, as well as from national governments. There is a significant overlap and agreement between organisations and researchers who have developed these principles for AI.
However, how teams implement these principles in the design and delivery of AI projects is more difficult. As situations and contexts differ, there is no ‘one-size fits all’ approach to implementation. While every government must take action towards trustworthy AI, exactly how they take action should be based on the structures, skills, and priorities of their government.
Developing an approach to implementing the principles of trustworthy AI in government projects is best begun by understanding the present position: how prepared is the government currently? What progress has already been made and where is there still work to be done? The Oxford Insights AI Readiness Assessment helps governments to answer these questions and determine their own existing strengths and weaknesses.
Who is it for?
The tool is intended for public servants working in AI policy and/or public sector innovation. The assessment will help public servants identify policy areas to focus on and also suggest routes forward; however the specific approach and actions your government opts to take will require further research and tailoring to the needs of your own government and people.
How the self-assessment tool works
The tool is structured around five pillars of trustworthy AI, based on the OECD principles for AI and adapted to the use of AI in the public sector.
The assessment is composed of a set of questions for each of the five pillars. Once all questions are answered the assessment provides users with a score from 0 to 100 for each pillar.
Start by downloading the Excel spreadsheet at the top of the page, and any answers and results you enter are only accessible by you.
This is our first version of the assessment. We are eager to hear from public servants who make use of the tool. Please reach out to us with feedback about your experience working through the assessment so we can improve both its content and design.
You are welcome to share your results with us and ask any questions. We warmly welcome any feedback on this first iteration of the tool. We will continue to update the tool in light of users’ experiences and reflections, and also as the AI governance field develops.
If you have suggestions for improving the tool, or to discuss results, please get in touch.