Just accountability structures – a way to promote the safe use of automated decision-making in the public sector
Open accessThe DOI matches its Crossref record; 40 of the 40 references it cites were found in a registry or an index; 12 of the 52 sources its claims cite were checked for support.
Abstract
The growing use of automated decision-making (ADM) systems in the public sector and the need to control these has raised many legal questions in academic research and in policymaking. One of the timely means of legal control is accountability, which traditionally includes the ability to impose sanctions on the violator as one dimension. Even though many risks regarding the use of ADM have been noted and there is a common will to promote the safety of these systems, the relevance of the safety research has been discussed little in this context. In this article, I evaluate regulating accountability over the use of ADM in the public sector in relation to the findings of safety research. I conducted the study by focusing on ongoing regulatory projects regarding ADM, the Finnish ADM legislation draft and the EU proposal for the AI Act. The critical question raised in the article is what the role of sanctions is. I ask if official accountability could mean more of an opportunity to learn from mistakes, share knowledge and compensate for harm instead of control via sanctions.
Overview
CoChat Summary
TL;DR
The article evaluates how legal accountability structures for automated decision-making (ADM) in the public sector align with safety research, focusing on Finland’s ADM legislation draft and the EU AI Act proposal. It argues that sanction-centered personal accountability may discourage error reporting and learning, while just-culture and organizational accountability could better support safe ADM use.
Key findings
- Finland’s draft ADM legislation emphasizes personal official accountability, requiring documentation and naming responsible persons for ADM development, introduction, and monitoring; this aims to ensure individuals can be held accountable but risks a blame culture and criminal sanctions.
- The EU AI Act proposal builds accountability around providers/users and administrative fines under Article 71, plus product-safety mechanisms such as CE marking and market withdrawal; incident reporting is limited to providers/users of high-risk systems and serious incidents/malfunctioning, with a high threshold and causal-link interpretation left to the reporting party.
- Safety research, especially the concept of just culture, recommends shifting from retributive to restorative justice and removing barriers to error reporting; clear duties and responsibilities are still valued, and serious or wilful misconduct can remain sanctionable.
- The paper concludes that safe ADM cannot be improved by focusing on personal accountability or sanctioning organizations alone; more attention should go to organizational culture, information sharing, and learning from mistakes.
Methods
- Normative legal analysis comparing accountability structures in two regulatory projects against findings from safety research literature.
- Materials: the Finnish government proposal HE 145/2022 vp and the EU AI Act proposal COM(2021) 206 final, including Presidency compromise texts dated 29.11.2021 and 15.6.2022; the material reflects the situation on 5 January 2023.
- Analysis focuses on drafted rules concerning accountability, sanctions, and information sharing, with a special focus on public-sector official accountability and sanctions.
Limitations
- Stated: one article cannot cover all aspects of safety and safe use of ADM; the focus is limited to the public sector, official accountability, and sanctions; the Finnish proposal could also be examined from other angles, but space limits this.
- Stated: the material is based on the situation on 5 January 2023; Finnish law drafting later continued, ended, and new regulations entered into force on 1 May 2023, while EU preparation continues, so the rules that actually came into force are not fully analyzed.
- Evident: the research does not separately include data collection; the analysis is conceptual/legal and does not empirically test the effects of accountability structures, and the AI Act analysis relies on draft and compromise texts rather than final law.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
10 Valid · 3 Not valid · 40 Not checked · 95 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“the public sector has certain significant features like being able to exercise public power which distinguish it from the private sector (Loi et al. 2021; Backes and Eliantonio 2017).”
§ Just accountability structures -a way to promote the safe use of automated decision-making in the public sector > 1.1 Digital catastrophe
What the sources say
- Not matchedcites Loi et al. 2021
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Supportedcites Backes and Eliantonio 2017 source
In “Various Instruments and Powers to Protect the General Interest,” the paper says that public-law competences are exercised exclusively by public authorities and cannot be held by private-law subjects. It gives raising taxes and issuing residence permits as examples, supporting the claimed distinction. (LLM, conf=0.93)
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.
- Not matched
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“one of the main points of the whole process is to clarify the official accountability carried personally by every civil servant and employee working with ADM (Suksi 2020).”
§ Just accountability structures -a way to promote the safe use of automated decision-making in the public sector > 4 ADM, personal accountability and sanctions
What the sources say
- Partly supportedcites Suksi 2020 sourceExperimental — accuracy not yet measured
Section 3.2 (p. 7) says Finland’s review of ADM regulation includes the liability of civil servants, and Section 5 (p. 13) says indirect liability would not suffice. These passages support clarifying personal accountability as an important issue, but do not say that every civil servant and employee working with ADM personally carries official accountability. (LLM, conf=0.82)
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.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“First, people responsible for overseeing and controlling the ADM system may act in ways that do not meet the system designer's expectations (Binns 2022).”
§ Just accountability structures -a way to promote the safe use of automated decision-making in the public sector > 4 ADM, personal accountability and sanctions
What the sources say
- Supportedcites Binns 2022 source
Section 5.1 says that people overseeing, interpreting, and reviewing algorithmic systems are unlikely to act in the ways designers expect. It also explains that their differing powers, pressures, motivations, and agendas can affect how they exercise judgment. (LLM, conf=0.94)
hedgedstates a possible cause- 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
“In addition, it has been argued that with human-technical systems, there are both human and machine contributions to decision making and thus, the false assumption that either a human or a machine must be at fault should be corrected (Wagner 2019).”
§ Just accountability structures -a way to promote the safe use of automated decision-making in the public sector > 4 ADM, personal accountability and sanctions
What the sources say
- Supportedcites Wagner 2019 source
In “Assumption of Binary Liability as a Challenge” (page unavailable), Wagner says that humans and technical systems both contribute to decision making and that a model requiring either a human or a machine to be at fault is unfit for such systems. This directly supports both parts of the citing sentence. (LLM, conf=0.95)
hedged- 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
“However, it is important to note that the proposed model is distinguished from pharmaceutical regulation, in which the authority performs the evaluation and grants permission to place the product on the market, as Michael Veale and Frederik Zuiderveen Borgesius have pointed out (2021).”
§ Just accountability structures -a way to promote the safe use of automated decision-making in the public sector > 5.1 ADM and accountability through sanctioning the organisation
What the sources say
- Supportedcites Veale and Zuiderveen Borgesius 2021 source
The section on the Draft AI Act and the New Legislative Framework (p. 6) explicitly contrasts the regimes: manufacturers generally carry out conformity assessment under the proposed model, whereas a public authority assesses pharmaceutical products before granting pre-marketing approval. This directly supports the attributed comparison. (LLM, conf=0.94)
plain- 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.
- Supported
41 more claims · 1 with their cited sources checked
Create a free account to see all 148 claims and to verify the 3 references still open.
- It may state two separate claims as one. 36
- Probably not a substantial claim. 87
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“According to Finnish law, official accountability is personally held by each public official engaged with ADM.”
From Reconciling Administrative Efficiency with Procedural Safeguards in the ADM System (2025) · cites this paper as Hirvonen, 2024
Citations
Cites
Analysing and Assessing Accountability: A Conceptual Framework 1
Bovens, M. (2007) · European Law Journal · doi:10.1111/j.1468-0386.2007.00378.x · Cited by 2,667 · cited 1×
“Sanctions are one inherent part of many interpretations of accountability (Bovens 2007; Lindberg 2013).”
Veale, M., & Zuiderveen Borgesius, F. (2021) · Computer Law Review International · doi:10.9785/cri-2021-220402 · Cited by 805
Kuziemski, M., & Misuraca, G. (2020) · Telecommunications Policy · doi:10.1016/j.telpol.2020.101976 · Cited by 610 · cited 2×
“Second, ADM systems that were originally designed to assist the human decision-maker may eventually have the final word in many cases, as it has been shown that civil servants using these systems are hesitant or unable to question the suggested outcomes (Kuziemski and Misuraca 2020; Wagner 2019).”
The Oxford Handbook of Ethics of AI
(2020) · doi:10.1093/oxfordhb/9780190067397.001.0001 · Cited by 459
Algorithmic Accountability and Public Reason
Binns, R. (2018) · Philosophy & Technology · doi:10.1007/s13347-017-0263-5 · Cited by 457 · cited 1×
“Reuben Binns handles the question of sufficient jus-tifications as one dimension of accountability and suggests that the notion of public reason could “answer to the problem of reasonable pluralism in the context of algorithmic decision making” (2018).”
Craig, P. (2006) · Oxford University Press eBooks · doi:10.1093/acprof:oso/9780199296811.001.0001 · Cited by 365 · cited 1×
“The requirement that the administration is both procedurally and sub-stantively accountable before the courts is central to the rule of law (Craig 2012).”
Wieringa, M. (2020) · Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency · doi:10.1145/3351095.3372833 · Cited by 321
The flaws of policies requiring human oversight of government algorithms
Green, B. (2022) · Computer Law & Security Review · doi:10.1016/j.clsr.2022.105681 · Cited by 263
The Flaws of Policies Requiring Human Oversight of Government Algorithms
Green, B. (2021) · doi:10.2139/ssrn.3921216 · Cited by 259
Mapping accountability: core concept and subtypes
Lindberg, S. I. (2013) · International Review of Administrative Sciences · doi:10.1177/0020852313477761 · Cited by 250 · cited 1×
“Sanctions are one inherent part of many interpretations of accountability (Bovens 2007; Lindberg 2013).”
Research works printed but not linked: 9
- Cobbe J, Lee SAM, Singh J (2021) Reviewable automated decision-making: a framework for accountable algorithmic systems. ACM Conference on Fairness, Accountability, and Transparency (FAccT ‘21), Virtual Event. Canada ACM, New York, pp 56–67
- Finck M (2019) Automated Decision-Making and Administrative Law (August 5, 2019) Max Planck Institute for Innovation & Competition Research Paper No. 19–10 https://ssrn.com/abstract=3433684. Also in the Oxford Handbook of Comparative Administrative Law, P. Cane et al. (eds.), Oxford, Oxford University Press, 2020, which has not been available to the author.
- Hirvonen H (2022) Virkavastuu ja päätösautomaatio – vastuun henkilökohtaisuus kriisissä? Lakimies 3–4(2022):386–418
- Jeffcott S, Weyman A, Pidgeon NF, Walls J (2006) Risk, trust, and safety culture in U.K. train operating companies. Risk Anal 26(2):1105–1121. https://doi.org/10.1111/j.1539-6924.2006.00819 · doi:10.1111/j.1539-6924.2006.00819
- Meijer A, Grimmelikhuijsen S (2020) Responsible and accountable algorithmization. How to generate citizen trust in governmental usage of algorithms. In: Schuilenburg M, Peeters R (eds) The Algorithmic Society: Technology, Power, and Knowledge, 1st edn. Routledge, Abingdon
- Waldman AE (2019) Power, process, and automated decision-making. Fordham L Rev 88:613–632
- Widlak A, van Eck M, Peeters R (2020) Towards principles of good digital administration. Fairness accountability and proportionality in automated decision-making. In: Schuilenburg M, Peeters R (eds) The Algorithmic Society: Technology, Power, and Knowledge, 1st edn. Routledge, Abingdon
- Journal of Safety Research webpage. A Safety and Health Research Forum. A Joint Publication of the National Safety Council http://www.nsc.org and Elsevier. https://www.journals.elsevier.com/journal-of-safety-research. Accessed 14 Aug 2022
- Safety Science webpage. Elsevier. https://www.sciencedirect.com/journal/safety-science. Accessed 15 Aug 2022
Cited by
Environmental stewardship in the oil and gas sector: Current practices and future directions
Simpa, P., Solomon, N. O., Adenekan, O. A., & Obasi, S. C. (2024) · International Journal of Applied Research in Social Sciences · doi:10.51594/ijarss.v6i5.1132 · Cited by 27
Calvano, M., Curci, A., Desolda, G., Esposito, A., Lanzilotti, R., & Piccinno, A. (2025) · Minds and Machines · doi:10.1007/s11023-025-09753-w · Cited by 33
Frontrunner model for responsible AI governance in the public sector: the Dutch perspective
Popa, D. M. (2024) · AI and Ethics · doi:10.1007/s43681-024-00596-2 · Cited by 13
“…as in cases where AI is used for police predicting, the child benefits scandal in the Netherlands that was characterized as a digital catastro-phe [1] and the recent case of the use of AI applications for handling of Visa requests by the Dutch Ministry of Foreign affairs (Tweede Kammer…”
Menke, W., Gomes, R. C., & Xavier, F. (2024) · Global Public Policy and Governance · doi:10.1007/s43508-024-00098-1 · Cited by 22
Christodoulou, P., & Limniotis, K. (2024) · Network · doi:10.3390/network4010005 · Cited by 15
“More recently, the Dutch tax authority had been wrongly accusing families of fraud, forcing them to repay benefits, due to the discriminatory problems arising from an algorithm in a decision-making system [22].”
Metadata
- Authors
- Hirvonen, H.
- Affiliations
- University of Helsinki (FI)
- Corresponding author
- Hirvonen, Hanne
- Venue
- AI & SOCIETY
- Published
- 2023-08-12
- Volume
- 39
- Issue
- 1
- Pages
- 155-167
- ISSN
- 0951-5666 (ISSN-L)
- 1435-5655
- OpenAlex
- W4385783856
- Semantic Scholar
- 4bf1525d76d5cc189cec7f93774813a9bd42601f
- Cited by
- 19
- References
- 37
- Citations by year
- 2026: 2 · 2025: 8 · 2024: 7 · 2023: 2
- FWCI
- 34.13
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- creativecommons.org/licenses/by/4.0 (tdm)
- creativecommons.org/licenses/by/4.0 (vor)
- Locations
- AI & Society (publishedVersion, cc-by)
- Työväentutkimus Vuosikirja (submittedVersion)
- Funders
- University of Helsinki including Helsinki University Central Hospital
- Topics
- Social Sciences › Law › Digitalization, Law, and Regulation
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Social Sciences › Law › Administrative Law and Governance
- Keywords
- automated decision-making, public sector, accountability, sanctions, AI regulation, EU AI Act, learning from mistakes, compensation for damages
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.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-07, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-07 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (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: Just accountability structures – a way to promote the safe use of automated decision-making in the public sector. https://app.cochat.ai/papers/10.1007/s00146-023-01731-z
- The paper
- Hirvonen, H. (2024). Just accountability structures – a way to promote the safe use of automated decision-making in the public sector. AI & SOCIETY. https://doi.org/10.1007/s00146-023-01731-z
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.