RESEARCH

Aoki’s recent research explores the societal implications of the transformation of the public sector by AI (AX), that is, the use of AI in the provision of government services and in government decision making. She conducts empirical research to generate evidence and stimulate discussion on these topics.

Over the past decade as a scholar, Aoki’s interest has centered on how the increasing adoption of AI in the public sector challenges the traditional field of public administration. In particular, she is intrigued by the possibility of a shift from “administrative discretion” to “AI discretion.” The former refers to human administrators exercising their own reasoning and judgment in processing tasks and making decisions, whereas the latter refers to AI autonomously making decisions and conducting operations. As AI becomes increasingly autonomous and agentic, AX in government could replace administrative discretion with AI discretion. Under such a redistribution of discretionary power, AX would not merely be about enhancing efficiency or convenience but would carry profound implications for the foundations of governance structures. Given this possibility, traditional theories premised on human agency have to be reexamined. At the same time, a new question emerges as to how we translate core human-centered administrative values—such as accountability, transparency, responsiveness, impartiality, and trustworthiness—to AI systems or socio-technical systems that combine human and AI agency. 

Recently Published Research 

Driven by her broad interest in normative issues surrounding public-sector AI, Aoki’s research has focused on specific, testable questions. Two of her most recent publications are introduced here.

① A Paper on “Zero Bureaucracy and Invisible Government”

This is a sole-authored article published in the Asia Pacific Journal of Public Administration. (The article is open access and can be read in full via the link below.)

Aoki, N., Tatsumi, T., Naruse, G., & Maeda, K. (2024). Explainable AI for government: Does the type of explanation matter to the accuracy, fairness, and trustworthiness of an algorithmic decision as perceived by those who are affected? Government Information Quarterly, 41(4), 101965. https://doi.org/10.1016/j.giq.2024.101965

In advanced countries, there has been a growing shift toward four types of services, which this paper refers to as “digital-era services”: (i) one-stop services, (ii) proactive services, (iii) virtual services, and (iv) services based on the once-only principle. The paper argues that these transformations are contributing to a shift toward “zero bureaucracy,” whereby citizens no longer need to take action to access government services, and “invisible government,” under which the interface between citizens and government effectively disappears. Although not fully discussed in the paper, AI technologies are expected to accelerate this transition by enhancing data interoperability. Increasingly autonomous and agentic AI could further bolster virtual services. 

Against this backdrop, the study used a conjoint analysis to examine Japanese citizens’ preferences regarding the transition from traditional administrative services to digital-era services. The analysis found that citizens showed the strongest preference for a shift toward the “once-only principle” over the traditional provision of services. The results also indicate a preference for a system that allows citizens to access both virtual and physical channels rather than one based exclusively on virtual operations. 

▶︎ An overview of this paper is also available on the faculty blog of the Graduate School of Public Policy.

② A Paper on Explainable AI (XAI)

This is a co-authored paper on explainable AI (XAI) in government, published in Government Information Quarterly. (The article is open access and can be read in full via the link below.)

Aoki, N., Tatsumi, T., Naruse, G., & Maeda, K. (2024). Explainable AI for government: Does the type of explanation matter to the accuracy, fairness, and trustworthiness of an algorithmic decision as perceived by those who are affected? Government Information Quarterly, 41(4), 101965. https://doi.org/10.1016/j.giq.2024.101965

Inspired by the European Union’s efforts to establish a right to explanations for certain AI-assisted decisions, this paper examines adverse algorithmic decisions made by government, focusing on tax audits and subsidy assessments. Specifically, it analyzes how differences in the type of explanation provided by a government affect how citizens subject to these decisions perceive their fairness, accuracy, and trustworthiness.

The paper argues that there is no single explanation that can fully and exhaustively capture the entire causal process underlying a decision (the explanandum). Any explanation is therefore inherently incomplete, representing only a selective “causal history.” Explanations can differ in the form, length, and level of abstraction, and these differences can be important factors shaping how citizens perceive and evaluate an algorithmic decision. This paper empirically demonstrated that citizens’ perceptions can vary depending on the type of explanation provided.

The key contribution of this study is to move beyond the question of whether an explanation should be provided at all and to draw attention to the type and nature of an explanation that is provided. It also highlights the need to consider not only what kinds of explanations are technically possible, but also what kinds of explanations are appropriate from the perspective of those receiving them.

▶︎ An overview of this paper is also available on the faculty blog of the Graduate School of Public Policy.

Other Research

In addition to work related to AI in the public sector, Aoki conducts research on other topics of public administration and management. Most recently, she published a systematic literature review in Public Administration on “whole-of-government” and “joined-up government” – concepts used to call for multi-agency collaboration to resist balkanization and siloed thinking. (The article is open access and can be read in full via the link below.)

Aoki, N., Tay, M., & Rawat, S. (2024). Whole-of-government and joined-up government: A systematic literature review. Public Administration, 102(2), 733–752. https://doi.org/10.1111/padm.12949

▶︎ An overview of this paper is also available on the faculty blog of the Graduate School of Public Policy.

Aoki’s publications since 2015, including those mentioned above, have appeared in the following journals:

  • The American Review of Public Administration
  • The Asian Journal of Political Science
  • Asia Pacific Journal of Public Administration
  • Cities
  • Computers in Human Behavior
  • Government Information Quarterly
  • Habitat International
  • The International Journal of Disaster Risk Reduction
  • The International Journal of Public Administration
  • The International Journal of Public Sector Management
  • The International Review of Administrative Sciences
  • Journal of Behavioral Public Administration
  • Local Government Studies
  • Public Administration
  • Public Administration and Development
  • Public Management Review
  • Risk Analysis
View Aoki’s Publication List