Inside algorithmic bureaucracy: Disentangling automated decision-making and good administration
Open accessThe DOI matches its Crossref record, with a small difference in title or year; 52 of the 52 references it cites were found in a registry or an index; 2 of the 11 sources its claims cite were checked for support.
Abstract
Public administrative bodies around the world are increasingly applying automated, administrative decision-making as underlying technologies such as machine learning mature. Such decision-making is a central element of emerging forms of algorithmic bureaucracies. With its direct exercise of public authority over individual citizens and firms, automated, administrative decision-making makes it particularly important to consider relations to values of good administration. Based on a multiple case-study, the article focuses on how empirical use of automated decision-making influences and transforms issues of good administration in four policy areas in Denmark: Business and social policy; labour market policy; agricultural policy; and tax policy. Supplementing emerging literature, the article exemplifies how public authorities struggle to apply automated decision-making in ways that support rather than undermine good administration. We identify six empirical relations of usage of automated, administrative decision-making and good administration: (I) Giving accurate and comprehensible reasons; (II) Informing addressees’ expectations; (III) Combining material and algorithmic expertise; (IV) Achieving effective oversight; (V) Continuously ensuring quality; and (VI) Managing high complexity. Additionally, we pinpoint related key capabilities for administrative bodies in order to support good administration.
Overview
CoChat Summary
TL;DR
This multiple case-study examines how automated, administrative decision-making (AADM) relates to good administration in four Danish policy areas. It identifies six empirical relations and six corresponding capabilities, showing AADM can both support and undermine good administration.
Key findings
- The paper identifies six empirical relations between AADM usage and good administration: giving accurate and comprehensible reasons; informing addressees’ expectations; combining material and algorithmic expertise; achieving effective oversight; continuously ensuring quality; and managing high complexity.
- Across the four cases, AADM usage had both supportive and undermining effects; all but one relation was observed in all cases, though context shaped whether effects were supportive or undermining.
- In cases 1, 3, and 4, 70–80% of administrative decisions were fully automated; case 2 used semi-automated decision-making where the system suggests procedural steps.
- The paper distils six key capabilities for administrative bodies, including transforming complex algorithms and high-volume data into comprehensible reasons, combining material and algorithmic expertise, ensuring continuous quality assurance, and managing “petite decisions” across wider algorithmic systems.
Methods
- Multiple case-study of four Danish policy areas: illness benefits, work retention, agricultural subsidies, and property value assessment, selected via purposive sampling for variation in professional and political complexity.
- Data collected October 2019–December 2020 through 60 interviews, 159 documents, and 13 observations across the cases.
- Thematic analysis using Boyatzis’ hybrid approach in NVivo, with within-case and cross-case analysis, producing 86 first-order topics, 18 second-order themes, and 6 aggregate dimensions.
Limitations
- The paper states it maintains an “inside-out” focus on administrative bodies and does not know whether the identified relations appear in data from citizens and firms subject to AADM.
- The paper states findings are context-specific and calls for future research on transfer to other policy areas and administrative traditions; it also notes no deterministic relations were found.
- Evident from the text: the study covers only Denmark and four cases, two decision-systems were under implementation, one relation in case 4 was not observed due to implementation stage, and access was not gained to a handful of potentially interesting documents.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
2 Valid · 7 Not valid · 9 Not checked · 105 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“AADM is a key element of emerging forms of 'algorithmic bureaucracy' (Vogl et al., 2020) where extensive use of technology heavily influences work practices, bureaucratic procedures, responsibilities of public servants, management practices, and organisational structures of public administrative bodies.”
What the sources say
- Supportedcites Vogl et al., 2020 source
The Discussion ('Findings suggest...' / closing of Discussion) defines algorithmic bureaucracy as an emergent form: 'These phenomena suggest a shift from traditional bureaucracy to algorithmic bureaucracy in local authorities' and 'The concept of algorithmic bureaucracy provides a framework to understand how computational algorithms are affecting all offices in the public sector, from the street to the system level,' accounting for automated processes made possible by algorithms, the roles/hierarchy/files of traditional bureaucracy, and their interaction. The Abstract adds that smart technologies 'add a new element to the socio-technical organization of public administration' and that algorithms 'may also change the nature of public administration,' matching the claim's list of affected work practices, procedures, responsibilities and structures; the paper's own caveat is that adoption is 'at an early, but foundational, stage' (nascent), so 'extensive use' is a slight strengthening rather than a mismatch. No passage contradicts the attribution. (LLM, conf=0.82)
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Particularly the former appears to be rooted in a somehow naïve belief in the quality of public servants' operation of decision-systems not taking into consideration behavioural issues such as automation bias (Cummings, 2006), inappropriate training or high work pressure.”
What the sources say
- Supportedcites Cummings, 2006 source
The citation is attached specifically to “automation bias,” not to the citing authors’ explanation about public servants or assessment schemes. In the passage preceding “Automation and Accountability,” Cummings describes people following incorrect automated recommendations despite opportunities to verify them and identifies automation bias as an important design consideration. (LLM, conf=0.90)
hedgedstates a possible causen = 4- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“On a digital map, fields are thus marked green, yellow, and red indicating whether the field is registered as complying with the basis of a given subsidy or not.”
plainSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The formal memoranda to farmers providing the reason for decisions on agricultural subsidies in case 3 thus often consists of 1015 pages of detailed text and maps making the reason de facto incomprehensible for the average addressee.”
plainstates a causeSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“If they do, it tends to be in an underdeveloped way, where initiatives are not systematically connected to a holistic approach to transparent decision processes.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
74 more claims
Create a free account to see all 123 claims and every source verdict.
- The quote only partly states this claim. 1
- It may state two separate claims as one. 27
- Probably not a substantial claim. 20
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 5. Managing relations of automated decision-making and good administration · p. 23
“Each relation manifests itself empirically in ways that both support and undermine good administration.”
§ 5. Managing relations of automated decision-making and good administration · p. 25
“[Example of undermining relation have not been observed.]”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 3.3 Data collection · p. 12
“Empirical data was collected in relation to each case in the form of interviews, documents, and observations in the period of October 2019 - December 2020 as described in Table 2:”
§ Supplemental Material: Appendix A (Overview of interviews) · p. 35
“August 28, 2019 December 20, 2019”
Citations
Cites
Qualitative Data Analysis: An Expanded Sourcebook
(1994) · Journal of Environmental Psychology · doi:10.1016/s0272-4944(05)80231-2 · Cited by 48,420
Qualitative research and evaluation methods
Patton, M. Q. (1980) · Cited by 35,733
(1998) · Cited by 13,501
A PUBLIC MANAGEMENT FOR ALL SEASONS
HOOD, C. (1991) · Public Administration · doi:10.1111/j.1467-9299.1991.tb00779.x · Cited by 10,159
The ethics of algorithms: Mapping the debate
Mittelstadt, B. D., Allo, P., Taddeo, M., Wachter, S., & Floridi, L. (2016) · Big Data & Society · doi:10.1177/2053951716679679 · Cited by 4,567 · cited 1×
“Administrative bodies ’ usage of AADM is strongly related to ongoing ethical and social debates on advanced technologies (e.g., Mittelstadt et al., 2016) including issues of data bias, threats to equity and lack of accountability.”
Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor
Lebovits, H. (2019) · Public Integrity · doi:10.1080/10999922.2018.1511671 · Cited by 3,279
Creating public value: strategic management in government
(1996) · Choice Reviews Online · doi:10.5860/choice.33-4147 · Cited by 2,552
Analyzing Qualitative Data: Systematic Approaches
Bernard, H. R., & Ryan, G. W. (2009) · Cited by 2,373
Jørgensen, T. B., & Bozeman, B. (2007) · Administration & Society · doi:10.1177/0095399707300703 · Cited by 1,082
Two Concepts of Accountability: Accountability as a Virtue and as a Mechanism
Bovens, M. (2010) · West European Politics · doi:10.1080/01402382.2010.486119 · Cited by 1,018
Research works printed but not linked: 8
- Cane P (2011) Administrative Law. 5th edition. Oxford: Oxford University Press, pp. 370–393.
- Danish National Audit Office (2013) Report on the Public Property Assessment (Summary). Copenhagen.
- European Commission (2021) eGovernment Benchmark 2021 - Insight Report. Brussels.
- Faculty of Social Sciences University Aalborg (2020) Guidelines for Promoting Responsible Research Practice. Aalborg. Available at: https://www.handbook.aau.dk/document?contentId=459534
- Margetts H (2021) The digital tools of government. Policy & Internet 13: 163–166.
- Misuraca G (2020) Rethinking democracy in the “pandemic society” A Journey in Search of the Governance with of and by AI. In: Proceedings of the 1st International Forum on Digital and Democracy Venice Italy 10–11 December 2020.
- Yin RK (2009) Case Study Research. 4th edition. Thousand Oaks, CA: Sage Publications.
- Young MM, Bullock JB, Lecy JD (2019) Artificial discretion as a tool of governance. Perspectives on Public Management and Governance 2(4): 301–313.
Cited by
Chandra, Y., & Feng, N. (2025) · Public Management Review · doi:10.1080/14719037.2025.2450680 · Cited by 87
Research on digital discretion—The subjectification of street‐level work in public administration
Johansson, J. (2025) · Review of Policy Research · doi:10.1111/ropr.12637 · Cited by 26
Automating street‐level discretion: A systematic literature review and research agenda
Northcott, E. (2025) · Policy Studies Journal · doi:10.1111/psj.70024 · Cited by 18
Wang, Y. (2025) · Organization Management Journal · doi:10.1108/omj-11-2024-2342 · Cited by 36
KAMU YÖNETİMİNDE ALGORİTMALARIN EGEMENLİĞİ: ALGOKRASİ VE TEHDİTLERİ
Polat, M. (2024) · Kamu Yönetimi ve Teknoloji Dergisi · doi:10.58307/kaytek.1495010 · Cited by 15
Metadata
- Authors
- Roehl, U., & Crompvoets, J.
- Affiliations
- Department of Digitalization, Copenhagen Business School, Frederiksberg, Denmark
- KU Leuven (BE)
- Venue
- Public Policy and Administration
- Publisher
- SAGE Publications
- Published
- 2023-08-31
- Volume
- 40
- Issue
- 2
- Pages
- 322-350
- ISSN
- 0952-0767 (ISSN-L)
- 1749-4192
- OpenAlex
- W4386307797
- Semantic Scholar
- e0bafae94521c646693be16809b292f53fe2c92a
- Cited by
- 21
- References
- 45
- Citations by year
- 2026: 12 · 2025: 5 · 2024: 4
- FWCI
- 20.62
- Type
- journal article
- Language
- en
- Open access
- bronze
- doi.org
- Licence
- journals.sagepub.com/page/policies/text-and-data-mining-license (tdm)
- Locations
- Public Policy and Administration (publishedVersion)
- CBS Research Portal (Copenhagen Business School) (acceptedVersion, other-oa)
- Lirias (submittedVersion, other-oa)
- Topics
- Social Sciences › Political Science and International Relations › E-Government and Public Services
- Social Sciences › Public Administration › Public Policy and Administration Research
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Keywords
- good administration, administrative transparency, explainability, public oversight, administrative expertise, quality assurance, labour market policy, tax policy, agricultural policy, social policy, Denmark
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Inside algorithmic bureaucracy: Disentangling automated decision-making and good administration. https://app.cochat.ai/papers/10.1177/09520767231197801
- The paper
- Roehl, U., & Crompvoets, J. (2025). Inside algorithmic bureaucracy: Disentangling automated decision-making and good administration. Public Policy and Administration. https://doi.org/10.1177/09520767231197801
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.
