JOP 88/4 pp. 1301-1317 2026-10-01 Original When Do Citizens Resist the Use of AI Algorithms in Public Policy? Theory and Evidence
Shir Raviv
Abstract
In recent years, there has been a significant rise in the use of algorithmic decision-making systems (ADS) to assist or replace human decision-making in a wide range of policy areas as policing, criminal sentencing, and social welfare assistance. How do citizens view the incorporation of this technology in guiding high-stakes decisions? I introduce a new theory to explain the conditions under which citizens view ADS as legitimate, fair, and accurate, and test it using data from original experiments embedded in a national U.S. survey. I show that across a wide range of policy domains, citizens exhibit aversion to the use of ADS in decisions that are seen as designed to sanction rather than to assist, and when they are required to make inferences about individuals rather than collectives. Evidence from a second experiment suggests that the employment of ADS in such contexts can significantly undermine the legitimacy of the policy actions they inform. Together, the study offers a framework to identify where AI-based tools will be deemed appropriate and where they might trigger backlash, highlighting the importance of accounting for citizens’ values in AI development and implementation to maintain legitimacy and democratic accountability.
JPART — 2026-10-01
Original Generative Spaces: Collaboration, Learning and Innovation in a Regulatory Sandbox
Carl Macrae, Christopher K Ansell
Abstract
Abstract Public administrators are increasingly called upon to create “generative spaces” that foster emergent processes of collaboration, learning and innovation. This article presents an inductive and ethnographically grounded study of one relatively novel but increasingly popular generative space—a regulatory sandbox. A regulatory sandbox is a controlled and defined space and set of processes that are established by regulators to work with innovators to collaboratively explore, test and learn about early-stage products and innovations and understand emerging regulatory implications. Investigating a sandbox to explore the regulation of new artificial intelligence technologies conducted by a national healthcare regulator, we identify the organizing practices of the sandbox that make it “generative” of collaboration, learning and innovation. An inductive analysis identified three key organizing practices: providing distinct and varied social spaces separated from normal regulatory demands (protection); enabling participants to collaboratively examine innovations and iteratively develop new knowledge and regulation (adaptation); and supporting focused and immersive examination of innovations from a diverse range of perspectives (participation). A theoretically-driven secondary analysis suggests that generative spaces can be understood as liminal experiences that foster detachment, equivalence and reconstruction.
PAR — 2026-10-01
Original Citizens' Expectations vs. Recruitment Signals: Are Public Administrations Trying to Attract the Ideal Public Servant?
Melissa Satzger Grøn, Dominik Vogel
PAR — 2026-09-30
Original Artificial Intelligence in Policing and Perceived Legitimacy: Does Gender Representation Make a Difference?
Canyu Gao, Norma M. Riccucci
Abstract
ABSTRACT As artificial intelligence (AI) increasingly transforms public administration, understanding its impact on perceived legitimacy becomes paramount, particularly, in high‐stakes areas such as policing. While existing research offers mixed evidence on whether government AI use enhances or undermines citizens' perceptions of fairness and trust, the conditions moderating these effects remain underexplored. Drawing on representative bureaucracy theory, this study uses a survey experiment to investigate how gender representation moderates the effect of AI use on citizens' perceived fairness and trust in domestic violence units (DVUs). Results show that female representation positively moderates the legitimacy impact of a DVU's AI use; whereas AI use decreases female respondents' perceived fairness and trust in male‐dominated DVUs, these negative effects disappear in female‐dominated DVUs. Our findings identify representation as a key context shaping citizens' responses to government AI use, extending representative bureaucracy theory by demonstrating its indirect symbolic effects.
PAR — 2026-09-29
Original Reply to Kettl: Beyond Bureaucracy: Why AI Demands a Fourth Way in Public Administration
Jack Baldwin, Tina Nabatchi
Abstract
ABSTRACT Kettl argues public administration faces a crisis rooted in widening gaps between scholars and practitioners and between the field's core ideas and current governance issues. We identify a critical example: artificial intelligence (AI) adoption in government. AI fundamentally transforms public organizations, redistributing tasks and discretion between humans and algorithms, embedding instrumental rationality into administrative processes, and reconfiguring authority structures. These developments challenge three longstanding pillars of public administration: bureaucratic neutrality, the politics‐administration dichotomy, and democratic governance. We argue that any proposed “Fourth Way” must center democratic resilience and technological stewardship, incorporating algorithmic transparency requirements and participatory governance structures that preserve public—and especially democratic—values.
PNAS 123/40 pp. e2613113123-e2613113123 2026-09-28 Original Classifying strategy use in multiattribute subjective choice: Application to conjoint experiments in political science
Nidhi V Banavar, Kirk C. Bansak
PSJ 54/4 2026-09-27
Original Problem and Solution Framing in Senate Tweets: Scaling Expert Interpretation With Machine Learning
Michelle M. Buehlmann, Misha Melnyk, Mitchell Dolny, Joshua D. Elkind, Alexander Michael Tjhin, Saisha Chebium, Blake VanBerlo, Annelise Russell, Jesse Hoey
PSJ 54/4 2026-09-27
Original Beyond Evidence: How Framing Shapes Public Health Policies During Health Crises
Antoine Claude Lemor, Eric Montpetit
Abstract
ABSTRACT Does the level of evidence that accompanies policy action depend on how a problem is framed, and on who frames it? This article examines these questions in the context of health crises. Bridging sociological and equivalence framing theories, it treats frames as preexisting interpretive schemes shared within communities of actors, whose downstream influence depends on the information available. Using a comprehensive textual database of COVID‐19 press conferences held in Quebec, where political decision‐makers held greater authority, and in Sweden, where scientists enjoyed more autonomy, the study combines natural language processing, OLS regressions, and Granger causality tests to analyze how frames and disclosed evidence were articulated with suppression and mitigation policies. Findings show that in Quebec, framing the pandemic as an imminent danger was associated with far‐reaching suppression measures even at low levels of disclosed evidence, while the moderate frame's association with mitigation strengthened as evidence increased. In Sweden, no relationship between framing and suppression policies appeared in public communications; framing instead operated jointly with evidence, the dangerous frame being associated with fewer mitigation references mainly at high levels of disclosed evidence. These results suggest that the relevant question is not how much evidence policy action requires, but by whom that evidence is required, and under what conditions. The evidentiary threshold attached to a frame appears to be a property of the community that carries it, with direct implications for how democratic institutions allocate influence between accountability and expertise.
PAR — 2026-09-27
Original Artificial Intelligence, Bureaucratic Discretion, and Democratic Administration
L. Jason Anastasopoulos
Abstract
ABSTRACT Since the start of the COVID ‐19 pandemic, many countries have been using artificial intelligence ( AI ) to replace civil service personnel. We argue that this practice undermines core democratic values. Using the public values framework, we identify four clusters of values that are threatened when bureaucratic discretion is transferred to algorithms: accountability, fairness and equity, individual consideration, and reason‐giving. Each of these values is incompatible with how algorithms render decisions. Because algorithms produce outputs based on statistical regularities, they diffuse responsibility across multiple parties, fail to make accurate predictions about anomalous cases, and attenuate citizens' claims to individualized consideration. Thus, this article frames AI adoption as a constitutional question about which decisions can be automated, and which should remain under human control because they implicate democratic values. We propose a centaur model in which AI supports rather than replaces human officials, which can preserve efficiency while protecting democratic governance.
PMR — pp. 1-24 2026-09-25 Original Role strain and job crafting in China’s digitalizing street-level bureaucracy
Fan Yang, Mingyang Yang, Shihan Lu, Yiyan Zhou, Junyao Zheng, Wenjie Li, Shaowei Liu