Format

P275. Building policy capacity for the AI era: how public servants and administrations learn to judge, use and govern artificial intelligence

State and Policy Capacity / Policy Design, Formulation & Instruments
PANEL CHAIR(S)
Ine Steenmans
University College London, University of London
Main chair
Maria Cross
University College London, University of London
Second chair
CATEGORISATION
POLICY TOPIC
State and Policy Capacity Policy Design, Formulation & Instruments
SECTOR
Administration and Civil Service Policy General Public Policy (no specific sector)
KEYWORDS
AI policy capacity Organisational learning
GENERAL OBJECTIVES, RESEARCH QUESTIONS AND SCIENTIFIC RELEVANCE

Public administrations are adopting artificial intelligence faster than they are building the capacity to use it well. Governments now expect public servants at every level to commission, interrogate, deploy and govern AI tools, and to make policy for societies increasingly shaped by AI. In response, substantial resources are flowing into AI literacy curricula, digital academies and leadership programmes. Yet we know little about which capabilities public servants actually need, whether and how learning initiatives build them, or whether these investments translate into broader institutional capabilities to use and govern AI.

 

Three literatures speak to this problem but to date little to each other. The policy capacity literature (Wu, Ramesh and Howlett 2015) distinguishes analytical, operational and political capacities at individual, organisational and systemic levels, but predates AI becoming part of everyday policy work; work on dynamic capabilities adds that organisational capacity is more than the sum of individual skills (Kattel and Mazzucato 2018). Public administration research on AI shows that it reshapes bureaucratic discretion and roles in ways that depend on organisational context (Bullock 2019; Giest and Klievink 2024), and that officials' use of algorithmic advice is shaped by automation bias, selective adherence and prior professional judgement (Alon-Barkat and Busuioc 2023; Selten, Robeer and Grimmelikhuijsen 2023). Technical skill alone is therefore insufficient. Research on public sector training, meanwhile, rarely tests whether learning changes behaviour or organisational performance rather than participants' reactions and knowledge, although the studies that do suggest it can, under particular conditions (Kroll and Moynihan 2015; Seidle, Fernandez and Perry 2016). The result is a gap between public investment in AI capability and our understanding of what that investment produces.

 

This panel asks: what capabilities does AI require of public servants and public organisations, and how do learning and development initiatives build them?

 

Papers may explore hypotheses such as:

 

  1.  AI capability in public administrations is primarily a matter of judgement rather than technical skill.

  2.  AI reconfigures existing policy capacities rather than adding a separate "digital" capacity.

  3.  Individual learning translates into organisational AI capability under particular conditions, such as permission to experiment and adequate data infrastructure.

  4.  Prevailing evaluation approaches cannot detect the capabilities that matter most, so new methods are needed to capture shifts in practice.

 

This panel's scientific relevance is threefold. Conceptually, it updates policy capacity theory for the AI era. Empirically, it brings together emerging evidence from AI literacy programmes, executive education, regulator training and on-the-job learning across diverse administrative contexts. Methodologically, it advances ways of evaluating capability development beyond satisfaction surveys. It will interest scholars of policy capacity, public sector innovation, digital governance and professional education, and practitioners who design and commission AI learning for government.

 

 

CALL FOR PAPERS

Governments are investing heavily in AI literacy, digital academies and leadership programmes, expecting public servants to commission, question, deploy and govern AI. Yet we know little about which capabilities this work actually demands, or whether learning initiatives build them. Is AI capability a matter of technical skill, or of judgement about when to trust, challenge or override a system? And how does individual learning become organisational capacity? This panel brings together research that tackles these questions.

 

We invite theoretical, empirical and methodological papers on how public servants and public organisations develop the capabilities to use, judge and govern AI. We welcome work from all administrative contexts and levels of government, and from both academic and practitioner authors.

 

Conceptual papers might, for example:

 

  •  theorise AI-related capabilities within, or against, existing policy capacity frameworks;

  •  examine the place of judgement, discretion, ethics and affect in AI-enabled policy work;

  •  critically analyse the assumptions behind AI literacy and digital skills agendas.

 

Empirical papers might, for example:

 

  •  identify what public servants, senior leaders or regulators need to know and be able to do in relation to AI;

  •  evaluate AI literacy programmes, digital academies, MPA/MPP curricula or executive education;

  •  study informal and on-the-job learning, such as experimentation, communities of practice or learning through deployment;

  •  compare national strategies for building AI capability in government;

  •  trace how individual learning does, or does not, become organisational capability.

 

Methodological papers might, for example:

 

  •  propose ways of evaluating capability development beyond reaction and knowledge measures;

  •  develop tools for assessing AI-related judgement in practice.

 

We particularly welcome papers that connect learning, capability and policy capacity theory, and that engage critically with the claims made for AI in government. Papers on AI as a policy instrument are welcome where they speak to the capacities of the people and organisations using it. Abstracts should state the research question, the capability or learning initiative studied, and the evidence or argument the paper will present.