The End of Advisors? Artificial Intelligence and the Transformation of Evidence-Based Policymaking

Recent developments in AI have promised a transformative change to the production and use of evidence by policymakers. The promise of AI in decision-making is undeniable – the pace, prediction and agility of decisions on an unprecedented scale. The emergence of new AI data and evidence synthesis tools has even led some to speculate we are nearing the end of scientific advisors as evidence and analysis are now at policymakers' fingertips. Nevertheless, it remains unclear whether AI in decision-making becomes a reality or another failed attempt at digitalising public services. The research so far has focused predominantly on the risks and promises of AI for decision-making rather than the everyday practices of experts, advisors and civil servants. This paper will start to address this gap by exploring the socio-technical transition of the science advisory landscape in the UK resulting from the introduction of AI evidence tools. Drawing on semi-structured interviews with experts and decision-makers, the paper will unpack three levels at which this transformation is enacted in practice: 1.) epistemics – including the impact of AI on epistemic justice and knowledge democracy; 2.) praxis – reconfiguration of roles and responsibilities resulting from AI use; 3.) institutions – emergent knowledge infrastructures around the AI and success and failures of different AI tools. Overall, this research offers a contribution to the scholarship on evidence and expertise in policymaking by focusing on the practices of governing with AI, rather than governing AI.