To create fairness and accountability in the use of government decision making algorithms

Conclusions and recommendations: The challenges of using data for good, of building fair and accountable systems for data driven decision making in government are numerous. Each of the experts I met with for this Fellowship had a different approach to a different part of this puzzle. The sheer number of dimensions on which experts in this field are working is dizzying. There are computationally focused solutions to the legal challenges, educationally focused solutions to the philosophical challenges, and socially focused solutions to the mathematical challenges. Hence, the four recommendations I give in this report are not a complete to-do list after which Government use of algorithmic decision making will be totally fair and accountable. Nor are they the only areas for improvement that I discovered throughout the course of this Fellowship. Rather, I chose these four recommendations because they were exceptional across a number of number of dimensions.

Each recommendation addresses an underlying theme from the Fellowship, creates systems that will be self sustaining and is likely to lead to further improvements. Most of all, each recommendation is ambitious. I am under no illusions about the difficulty of implementing any of these recommendations. However, I am dedicated to investing my energy and recruiting others to help bring them about. Finally, I do not propose these recommendations as a means to criticise the status quo. There are wonderful, capable, bright and hard working data science professionals in the public sector of South Australia. They are supported by strong systems that help data flow and provide exceptional oversight. I humbly propose these recommendations because things can always be better. Please see my report for my proposed four ways to help create fairness and accountability in the use of government decision making algorithms in South Australia.

Keywords: Artificial intelligence, ethics, government, regulation, decision making