The Future of Good Decisions: An Evolutionary Approach to Human-AI Government Administrative Decision-Making

Administrative decision-making is a core aspect of government, affecting our daily lives. Decisions, such as assessing visa applications or calculating social security entitlements, are regulated by law that ensures decision-makers use their power lawfully and fairly. This reflects 'rule of law' values, aiming to make governance accountable, fair and transparent. At the practical level, these values are upheld by 'good decisions' protocols that guide individual decision-makers to be impartial, thorough and rational.

But decision-making is undergoing a 'quiet revolution'. As Artificial Intelligence (AI) is increasingly used to make administrative decisions more effective in a technologizing world, concern is growing over its incompatibility with the rule of law. The scale, speed, complexity and decentralisation of big data, neural networks and algorithms disrupt the central procedural expectations of human decision-making and current 'good decisions' guidance, being difficult to examine with the human brain. Work is underway to reshape some aspects of AI in government, to minimise bias and maximise transparency. But leading scholars suggest a deeper problem: our rule of law values are incompatible with algorithmic technology at a basic level-yet these are the only standards we have for evaluating government decisions.

This research will ask: how can our ideas of 'good decisions' evolve for a government where humans and AI increasingly 'co-decide'? Legal systems have seen radical change in the past; absorbing new technologies like the printing press, or modern scientific reasoning. Good government has often involved asking what it means to be human, and how to design institutions accordingly. But today, these problems have a different complexion. The arrival of AI makes it urgent to rethink not only the quality of reasoning in decisions, but the basic idea of decisions as the product of individual human thought, immune from the technological and material sphere. This will be ever less sustainable as the physical, digital and biological continue to blur. What will be the next evolution in the way we make and evaluate decisions, for the good governance of a new 'techno-social' life?

This fellowship will unite the interdisciplinary knowledge and methods necessary to deeply rethink and redesign the central notion of 'good decisions', for a coming age without clear oppositions between humans and technology, law and designed systems. Understanding AI's disruption to deliberative governance as evolutionary change, it embraces the need for a new 'ecological' view of law as emerging from social networks, and connected to fundamental principles of life. It will enlist rich literatures from the philosophy of evolution, legal theory and critical AI studies to retheorise decision-making, and leverage practical approaches in the creative sphere to explore new procedural and technological models for decisions. It will lead experts from across law, science, government and the humanities to test and collaboratively remake these models, using 'prefigurative' 'Live Action Role Play' methods to learn how decision design, regulation and evaluation might work in a networked, digital public. Making a major advance in legal theory, this research will underpin practical change by carefully rethinking decision-making's normative foundations. It will integrate its insights into contemporary legal doctrine and government practice with guidance from expert lawyers and judges, leading the way to concrete improvements in governance for the digital age.

This fellowship will lead the coordinated intellectual, creative, collaborative and practical change that is necessary to confront the challenge - and opportunity - that AI presents for governance. It will tell a compelling new story about the co-evolution of humans, law and technology, and equip a new generation of thinkers to work with an integrated view of law, science, technology and culture.