AI and the Transformation of Accountability and Discretion in Urban Governance

Juncheng "Tony" Yang; Stephen Goldsmith

arxiv: 2502.13101 · v4 · submitted 2025-02-18 · 💻 cs.CY

AI and the Transformation of Accountability and Discretion in Urban Governance

Stephen Goldsmith , Juncheng "Tony" Yang This is my paper

Pith reviewed 2026-05-23 02:41 UTC · model grok-4.3

classification 💻 cs.CY

keywords AI governanceurban governancebureaucratic discretionaccountabilityalgorithmic systemspublic administrationcitizen participationdata bias

0 comments

The pith

AI redistributes discretion and accountability across institutional levels, roles, and citizen interactions in urban governance rather than simply restricting or enhancing them.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper claims that artificial intelligence in city operations shifts where and how decisions get made instead of uniformly cutting back or increasing flexibility for front-line staff. This shift influences political oversight by elected leaders, professional standards among administrators, and participatory input from residents. The authors introduce accountable discretion as a framework and tie it to five principles meant to guide implementation. These include ensuring equal access to AI tools, building adaptive structures, strengthening data governance, keeping humans in key decisions, and involving citizens in oversight. The argument matters because it treats the balance between judgment and responsibility as a changing relationship shaped by technology and institutions rather than a fixed trade-off.

Core claim

The paper argues that AI does not simply restrict or enhance discretion but redistributes it across institutional levels, professional roles, and citizen interactions. This redistribution affects political, professional, and participatory forms of accountability while introducing risks such as data bias, algorithmic opacity, and fragmented responsibility. In response, the study proposes the concept of accountable discretion and links it to guiding principles with actionable measures: equal AI access, adaptive administrative structures, robust data governance, proactive human-led decision-making, and citizen-engaged oversight. The work reframes the discretion-accountability relationship as a

What carries the argument

Redistribution of discretion, which carries the argument by moving decision authority between managers, professionals, and citizens rather than eliminating or expanding it at a single level.

If this is right

Managerial oversight strengthens through AI-enabled monitoring of service delivery.
Service consistency improves because algorithms apply rules more uniformly across cases.
Citizen access to information expands as AI systems make government data more reachable.
Accountability becomes more fragmented when responsibility for outcomes spreads across human and algorithmic actors.
The traditional tension between discretion and accountability turns into a dynamic relationship that institutions must actively manage.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The redistribution logic could apply to AI use in non-urban public services if the institutional mechanisms prove similar.
Cities could test the proposed principles by comparing decision patterns in departments with and without the five measures in place.
Training programs for public administrators may need updates to prepare staff for roles that emphasize oversight of algorithmic outputs rather than direct case handling.

Load-bearing premise

Illustrative cases from urban settings are sufficient to establish a general redistribution of discretion and accountability without systematic empirical testing or quantitative validation.

What would settle it

Longitudinal data from multiple cities tracking the distribution of decision authority and accountability metrics before and after AI deployment that shows no measurable shift across levels would undermine the central claim.

read the original abstract

This paper offers a conceptual analysis of the transformative role of Artificial Intelligence (AI) in urban governance, focusing on how AI can reshape the relationship between bureaucratic discretion and accountability. Drawing on public administration theory and algorithmic governance research, the study argues that AI does not simply restrict or enhance discretion but redistributes it across institutional levels, professional roles, and citizen interactions. While primarily conceptual, this paper uses illustrative cases to show that AI can strengthen managerial oversight, improve service delivery consistency, and expand citizen access to information. These changes affect different forms of accountability: political, professional, and participatory, while introducing new risks, such as data bias, algorithmic opacity, and fragmented responsibility across actors. In response, the paper introduces the concept of accountable discretion and proposes guiding principles, each linked to actionable measures: equal AI access, adaptive administrative structures, robust data governance, proactive human-led decision-making, and citizen-engaged oversight. This study contributes to the AI governance literature by moving beyond narrow concerns with perceived discretion at the street level, highlighting instead how AI transforms rule-based discretion across governance systems. It also reframes the trade-off between discretion and accountability as a dynamic and evolving relationship shaped by algorithmic systems and institutional practices.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper reframes AI as redistributing discretion in urban governance rather than restricting or enhancing it, but the claim rests on conceptual synthesis and illustrative cases without empirical tests.

read the letter

The main takeaway is that this paper synthesizes public administration theory with algorithmic governance work to argue AI shifts discretion across institutional levels, roles, and citizen interactions instead of simply limiting or expanding it. It coins accountable discretion and ties it to five principles with measures like equal access and citizen oversight. That framing organizes some existing ideas about accountability types and risks such as bias and opacity in a single place, which could help readers in public admin track how AI changes rule-based decisions. The literature connections are straightforward and the illustrative urban cases show concrete ways managerial oversight or service consistency might shift. The central redistribution claim, however, is asserted through those cases rather than derived from a mechanism or checked against data, so it functions more as a hypothesis than a demonstrated pattern. No counterexamples or systematic comparisons appear, leaving open whether restriction or enhancement remains the more common outcome in practice. This is a paper for public administration scholars and AI governance readers who already work in the conceptual space; it will not supply new data, models, or falsifiable predictions for technical audiences. The thinking is coherent on its own terms and engages the cited literature honestly, even if the novelty is mostly extension rather than a fresh result. It deserves peer review at a venue that publishes conceptual bridging pieces, though the authors would need to address the evidence gap for broader impact.

Referee Report

2 major / 2 minor

Summary. This paper offers a conceptual analysis arguing that AI reshapes the relationship between bureaucratic discretion and accountability in urban governance by redistributing discretion across institutional levels, professional roles, and citizen interactions, rather than simply restricting or enhancing it. Drawing on public administration theory and algorithmic governance research, it uses illustrative cases to demonstrate impacts on political, professional, and participatory accountability, introduces risks such as data bias and opacity, and proposes the concept of 'accountable discretion' with five guiding principles and linked actionable measures.

Significance. The paper's reframing of the discretion-accountability trade-off as dynamic and evolving under AI systems contributes to the AI governance literature by shifting focus from narrow street-level concerns to systemic transformations. The conceptual synthesis and proposal of guiding principles for 'accountable discretion' provide a valuable framework for future research and policy, particularly if supported by further empirical validation. The strength lies in the integration of theory with illustrative examples to highlight both opportunities and risks in AI deployment for urban governance.

major comments (2)

[Abstract] Abstract: The central claim that AI 'does not simply restrict or enhance discretion but redistributes it' across levels, roles, and citizen interactions is load-bearing for the contribution beyond prior literature, yet it rests on illustrative cases used as existence proofs without a derived mechanism predicting redistribution as the dominant outcome or analysis of counterexamples where restriction or enhancement prevails.
[Illustrative cases and proposal of accountable discretion] Section on illustrative cases and accountable discretion principles: The five guiding principles (equal AI access, adaptive administrative structures, robust data governance, proactive human-led decision-making, citizen-engaged oversight) and their measures presuppose the general redistribution pattern; the cases illustrate specific benefits like strengthened managerial oversight and service consistency but do not provide systematic evidence distinguishing redistribution from net restriction or enhancement across governance systems.

minor comments (2)

[Abstract] The abstract refers to 'five guiding measures' without enumerating them, which reduces immediate clarity for readers before reaching the later proposal.
[Introduction and conceptual framework] The introduction of 'accountable discretion' as a new concept would benefit from explicit contrasts with related terms in the cited public administration and algorithmic governance literature to avoid potential overlap.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which help clarify the boundaries of our conceptual contribution. We agree that the central claim requires more explicit support and will revise the manuscript accordingly while preserving its primarily conceptual nature.

read point-by-point responses

Referee: [Abstract] Abstract: The central claim that AI 'does not simply restrict or enhance discretion but redistributes it' across levels, roles, and citizen interactions is load-bearing for the contribution beyond prior literature, yet it rests on illustrative cases used as existence proofs without a derived mechanism predicting redistribution as the dominant outcome or analysis of counterexamples where restriction or enhancement prevails.

Authors: We accept this critique. The paper draws on public administration theory to posit redistribution as a primary dynamic, but the abstract and main text do not sufficiently articulate mechanisms or counterexamples. In revision we will (1) add a short subsection outlining mechanisms such as data centralization shifting discretion upward and (2) include a brief discussion of counterexamples (e.g., fully automated permitting systems that net-restrict discretion). The abstract will be updated to reflect this added nuance. revision: yes
Referee: [Illustrative cases and proposal of accountable discretion] Section on illustrative cases and accountable discretion principles: The five guiding principles (equal AI access, adaptive administrative structures, robust data governance, proactive human-led decision-making, citizen-engaged oversight) and their measures presuppose the general redistribution pattern; the cases illustrate specific benefits like strengthened managerial oversight and service consistency but do not provide systematic evidence distinguishing redistribution from net restriction or enhancement across governance systems.

Authors: We agree the principles are framed around redistribution and that the cases function as illustrations rather than systematic comparisons. Revision will expand the cases section with explicit analysis showing how each example redistributes rather than merely restricts or enhances discretion, and will add a limitations paragraph noting the absence of cross-system empirical tests. The principles themselves will remain normative but will be more clearly tied to the redistribution argument. revision: partial

Circularity Check

0 steps flagged

No circularity in conceptual synthesis from external literature

full rationale

The paper advances a conceptual argument by synthesizing public administration theory and algorithmic governance research, supported by illustrative urban cases. No equations, parameter fitting, or self-definitional reductions appear. The redistribution claim is framed as an interpretive synthesis rather than a derivation that collapses to its own inputs. Cited literature is external; no load-bearing self-citation chains or uniqueness theorems imported from the authors' prior work are present. Illustrative cases function as examples, not as fitted inputs renamed as predictions. The derivation chain remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The analysis rests on domain assumptions from public administration theory and introduces one new conceptual entity without independent evidence or data.

axioms (2)

domain assumption Public administration theory and algorithmic governance research provide an appropriate foundation for analyzing AI effects on discretion and accountability.
Invoked in the abstract to frame the entire conceptual analysis.
ad hoc to paper Illustrative cases can demonstrate systemic transformations in accountability forms.
Used to support arguments about redistribution and risks.

invented entities (1)

accountable discretion no independent evidence
purpose: A proposed concept to balance AI-enabled discretion with accountability mechanisms in governance.
Introduced as a response to identified risks of bias, opacity, and fragmented responsibility.

pith-pipeline@v0.9.0 · 5741 in / 1211 out tokens · 30008 ms · 2026-05-23T02:41:41.021193+00:00 · methodology

discussion (0)

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Governing What the EU AI Act Excludes: Accountability for Autonomous AI Agents in Smart City Critical Infrastructure
cs.CY 2026-05 unverdicted novelty 5.0

The EU AI Act narrows accountability for multi-agent AI in critical infrastructure by excluding safety components from key explanation and impact assessment rights, and the paper proposes AgentGov-SC, a three-layer ar...

Reference graph

Works this paper leans on

4 extracted references · 4 canonical work pages · cited by 1 Pith paper

[1]

Stanford Center for Research on Foundation Models (CRFM)

The Foundation Model Transparency Index. Stanford Center for Research on Foundation Models (CRFM). Boulus-Rødje, N., Cranefield, J., Doyle, C., Fleron, B., 2024. GenAI and me: the hidden work of building and maintaining an augmentative partnership. Pers Ubiquit Comput 28, 861–874. https://doi.org/10.1007/s00779-024-01810-y Bovens, M., 1998. The quest for ...

work page doi:10.1007/s00779-024-01810-y 2024
[2]

State Agent or Citizen Agent: Two Narratives of Discretion

https://doi.org/10.1093/jopart/mum030 Maynard-Moody, S., Musheno, M., 2000. State Agent or Citizen Agent: Two Narratives of Discretion. Journal of Public Administration Research and Theory 10, 329 –358. https://doi.org/10.1093/oxfordjournals.jpart.a024272 Meijer, A., Lorenz, L., Wessels, M., 2021. Algorithmization of Bureaucratic Organizations: Using a Pr...

work page doi:10.1093/jopart/mum030 2000
[3]

The Hype Is Real

https://doi.org/10.1111/1467-9299.00218 Mulgan, R., 2003. Issues of Accountability, in: Mulgan, R. (Ed.), Holding Power to Account: Accountability in Modern Democracies. Palgrave Macmillan UK, London, pp. 1 –35. https://doi.org/10.1057/9781403943835_1 36 Mulligan, D.K., Bamberger, K.A., 2019. Procurement As Policy: Administrative Process for Machine Learn...

work page doi:10.1111/1467-9299.00218 2003
[4]

An MIT Exploration of Generative AI

People-Powered Gen AI: Collaborating with Generative AI for Civic Engagement. An MIT Exploration of Generative AI. https://doi.org/10.21428/e4baedd9.f78710e6 38 Wirtz, B.W., Weyerer, J.C., Geyer, C., 2019. Artificial Intelligence and the Public Sector — Applications and Challenges. International Journal of Public Administration 42, 596–615. https://doi.or...

work page doi:10.21428/e4baedd9.f78710e6 2019

[1] [1]

Stanford Center for Research on Foundation Models (CRFM)

The Foundation Model Transparency Index. Stanford Center for Research on Foundation Models (CRFM). Boulus-Rødje, N., Cranefield, J., Doyle, C., Fleron, B., 2024. GenAI and me: the hidden work of building and maintaining an augmentative partnership. Pers Ubiquit Comput 28, 861–874. https://doi.org/10.1007/s00779-024-01810-y Bovens, M., 1998. The quest for ...

work page doi:10.1007/s00779-024-01810-y 2024

[2] [2]

State Agent or Citizen Agent: Two Narratives of Discretion

https://doi.org/10.1093/jopart/mum030 Maynard-Moody, S., Musheno, M., 2000. State Agent or Citizen Agent: Two Narratives of Discretion. Journal of Public Administration Research and Theory 10, 329 –358. https://doi.org/10.1093/oxfordjournals.jpart.a024272 Meijer, A., Lorenz, L., Wessels, M., 2021. Algorithmization of Bureaucratic Organizations: Using a Pr...

work page doi:10.1093/jopart/mum030 2000

[3] [3]

The Hype Is Real

https://doi.org/10.1111/1467-9299.00218 Mulgan, R., 2003. Issues of Accountability, in: Mulgan, R. (Ed.), Holding Power to Account: Accountability in Modern Democracies. Palgrave Macmillan UK, London, pp. 1 –35. https://doi.org/10.1057/9781403943835_1 36 Mulligan, D.K., Bamberger, K.A., 2019. Procurement As Policy: Administrative Process for Machine Learn...

work page doi:10.1111/1467-9299.00218 2003

[4] [4]

An MIT Exploration of Generative AI

People-Powered Gen AI: Collaborating with Generative AI for Civic Engagement. An MIT Exploration of Generative AI. https://doi.org/10.21428/e4baedd9.f78710e6 38 Wirtz, B.W., Weyerer, J.C., Geyer, C., 2019. Artificial Intelligence and the Public Sector — Applications and Challenges. International Journal of Public Administration 42, 596–615. https://doi.or...

work page doi:10.21428/e4baedd9.f78710e6 2019