Abstract
This chapter deals with the highly pertinent question of whether the use of AI technologies by a government agency in performing its administrative tasks would constitute delegation of state powers. This question is analyzed against the backdrop of the CE and the administrative-law framework. In conjunction with this, the chapter also highlights certain AI use cases that would not constitute delegation of state power. The aim behind scrutinizing these use cases is to shed light on the underlying question of the key considerations for ascertaining whether and to what extent an administrative task could be delegated to AI.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Change history
23 February 2024
A correction has been published.
Notes
- 1.
Such a principle emerges in light of the third paragraph of the Constitution of the Republic of Estonia. See also § 54 APA, whereby an administrative act is lawful if it is issued by a competent administrative authority pursuant to legislation in force at the moment of the act.
- 2.
See § 4(1) APA, which foresees the right of discretion consisting of authorization granted specifically to an administrative authority. An administrative act is lawful if issued by a competent administrative authority, § 54(1) APA.
- 3.
See also Merusk et al. (2012), comments 2.1.3 and 2.3.1.1.
- 4.
The actions of an individual in a particular situation (e.g., choosing the speed of the scooter) are limited with reference to what the technology enables.
- 5.
See Coglianese and Lehr (2017).
- 6.
Such deterministic systems are often referred to as legal expert systems. See, for example, Greenleaf et al. (2018).
- 7.
On the functioning of AI and ML, see also Sect. 2.1.
- 8.
See Ashely (2017), p. 107.
- 9.
Command and control are the two universal characteristics of regulation—see Leenes (2011), pp. 147–148.
- 10.
See also Coglianese and Lehr (2017), p. 1176: “[M]achine learning algorithms have the potential to transform key governmental functions in ways that not only augment human judgment but replace it with automated, algorithmic analysis.”
- 11.
Ibid.
- 12.
See Narits et al. (2017), comment 6.
- 13.
Or, at the very least, less likely to be corrupted. See Coglianese and Lehr (2017), pp. 1180–1181.
- 14.
Ibid. But see also Sect. 5.2.2.2, concerning the possible danger of an “over-reliance” on AI solutions.
- 15.
See Parrest (2009), pp. 48–50.
- 16.
- 17.
See Madise (2020), comment 3.
- 18.
See also Bayamlıoğlu and Leenes (2018), p. 32.
- 19.
Ibid.
- 20.
On the risk of AI’s opaqueness potentially undermining the will of the legislator, see the discussioin in Sect. 6.6.
- 21.
See Bayamlıoğlu and Leenes (2018), p. 29.
- 22.
See also Pilving and Mikiver (2020a), p. 55 et seq.
- 23.
- 24.
See also AI HLEG (2019), pp. 14 et seq.
- 25.
For example, as many as 70,000 unemployed people in Sweden were wrongly denied benefits by a government entity because 10–15% of the decisions made by an automated system were erroneous; see Wills (2019).
- 26.
See Art. 3 of the Constitution of the Republic of Estonia and Parrest (2009), p. 52.
- 27.
Parrest (2009), p. 50.
- 28.
See Hildebrandt (2017), p. 10.
- 29.
See § 1 APA.
- 30.
According to the APA, some of these are carrying out administrative acts (§ 51 APA), executing administrative measures (§ 106 APA), issuing administrative regulations (§ 88 APA) and concluding administrative contracts (§ 95 APA).
- 31.
Administrative acts may be issued to regulate individual cases in a public-law relationship; see § 51(1) and (2) APA.
- 32.
§ 106 APA.
- 33.
Ibid; see also Narits et al. (2017), comment 19 et seq.
- 34.
See § 4(1) APA; the right of discretion is foreseen as an authorization granted specifically to an administrative authority.
- 35.
See also Merusk et al. (2012), comments 2.1.3 and 2.3.1.1.
- 36.
See also Parrest (2009), pp. 51–52.
- 37.
Coglianese and Lehr (2017), p. 1178.
- 38.
Riigikohtu üldkogu otsus (Decision of the Estonian Supreme Court en banc) (2007) 3-3-1-41-06, section 21 and Riigikohtu üldkogu otsus (Decision of the Estonian Supreme Court en banc) (2008) 3-4-1-19-07, section 25.
- 39.
For instance, the City of Chicago uses machine learning to support its services through its SmartData Platform, to assess what restaurants should be inspected and evaluate where and when rodent-control bait should be used; see Finck (2019), p. 6.
- 40.
For example, the Estonian Expobot (see https://www.expo2020.ee/?lang=en) takes advantage of the power of AI to learn to answer questions about Estonia in all languages. However, it must be noted that a chatbot could also be combined with a decision-making system that, for instance, grants access to further services on the basis of the information exchanged via the chatbot. See also Martini et al. (2021), p. 3.
- 41.
See §§ 107-109 APA.
- 42.
See § 107(2) APA.
- 43.
See also Aedmaa et al. (2004), pp. 476, 469–470; § 107(1) APA.
- 44.
See Pilving and Mikiver (2020b), § 3.3.
- 45.
See, for example, Hildebrandt (2018), p. 2.
- 46.
In essence, the prediction provided by such AI-based support tools functions as a cognitive bias, which conditions the user toward believing the prediction when there is no other tangible metric available for assessing a fair outcome. The science behind anchoring heuristics is most famous from the theorizing of Tversky and Kahneman (1974). See also Chen (2019); Stein and Drouin (2017); Cummings (2004).
- 47.
See Pilving and Mikiver (2020b), § 3.3.
- 48.
See, for example, § 462 TA; § 336 ECA; § 4892 CoCP. See also Sect. 3.3.
- 49.
See § 35a of the German VwVfG. This paragraph came into force on 1 January 2017 in connection with the legal amendment “Gesetz zur Modernisierung des Besteuerungsverfahrens,” Vol. 18 July 2016, BGBl. I, p. 1679 (No. 35).
- 50.
- 51.
- 52.
- 53.
Martini and Nink (2018), p. 1128.
- 54.
Beschlussempfehlung und Bericht des Bundestags-Finanzausschusses zu dem Gesetzentwurf der Bundesregierung – Drucksache 18/7457 – BT-Drs. 18/8434, p. 122; Stelkens (2018), para. 1.
- 55.
Stegmüller (2018), p. 353.
- 56.
- 57.
- 58.
Compare: APA Proposal (2022). Annex to the explanatory memorandum (comments and suggestions), p. 1.
- 59.
§ 71 (2)(2) APA Proposal (2022).
- 60.
§ 71 (2)(3) APA Proposal (2022).
- 61.
§ 71 (3)(1) APA Proposal (2022).
- 62.
§ 71 (3)(2) APA Proposal (2022).
- 63.
§ 71 (3)(1) APA Proposal (2022).
- 64.
§ 71 (3)(5) APA Proposal (2022).
- 65.
§ 71 (2)(5) and (6) APA Proposal (2022).
- 66.
It should, however, be noted that such AI systems may be regulated to some degree by the AIA (most notably if they are considered “high-risk systems” or are subject to the transparency requirements of Art. 52 AIA Proposal), which simultaneously establishes specific qualitative and procedural requirements for the AI systems while also restricting the possibility of enacting any further requirements.
- 67.
See also: Ministry of Justice of Estonia (2022), p. 6 et seq.
- 68.
Ibid.
- 69.
Riigikohtu halduskolleegiumi otsus (Estonian Supreme Court Administrative Law Chamber decision) (2008) 3-3-1-81-07, section 13.
- 70.
Ibid., section 14. Translation by the authors.
- 71.
Cf. also: Ministry of Justice of Estonia (2022), p. 7.
- 72.
Voßkuhle and Kaufhold (2016), pp. 315–316.
- 73.
See above and § 71 (3)(1) APA Proposal (2022).
- 74.
ERR – Estonian Public Broadcasting. “Estonia’s new government takes office,” 18 July 2022, https://news.err.ee/1608660520/estonia-s-new-government-takes-office.
- 75.
See also APA Proposal (2022) and Ministry of Justice of Estonia (2022).
References
Aedmaa A et al (2004) Haldusmenetluse käsiraamat. Tartu Ülikooli Kirjastus, Tartu
AI HLEG (2019) Ethics guidelines for trustworthy AI (Ethics Guidelines). Brussels, 8 April 2019. https://ec.europa.eu/newsroom/dae/document.cfm?doc_id1/458477https://ec.europa.eu/futurium/en/ai-alliance-consultation
Ashely KD (2017) Artificial intelligence and legal analytics: new tools for law practice in the digital age. Cambridge University Press
Bayamlıoğlu E, Leenes R (2018) The “rule of law” implications of data-driven decision-making: a techno-regulatory perspective. Law Innov Technol 10(2):295–313. https://doi.org/10.1080/17579961.2018.1527475
Chen DL (2019) Machine learning and the rule of law. In: Livermore MA, Rockmore DN (eds) Law as data: computation, text & the future of legal analysis. Santa Fe Institute Press, pp 433–442
Coglianese C, Lehr D (2017) Regulating by robot: administrative decision making in the machine-learning era. Georgetown Law J 105:1147. https://ssrn.com/abstract1/42928293
Cummings ML (2004) Automation bias in intelligent time critical decision support systems . AIAA 1st Intelligent Systems Conference, Chicago USA, September 2004
Finck M (2019) Automated Decision-Making and Administrative Law. Max Planck Institute for Innovation and Competition Research Paper No. 19-10 https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3433684
Greenleaf G et al (2018) Building sustainable free legal advisory systems: experiences from the history of AI & law. Comput Law Secur Rev 34(2). https://doi.org/10.1016/j.clsr.2018.02.007
Hildebrandt M (2017) Law as computation in the era of artificial legal intelligence: speaking law to the power of statistics. https://doi.org/10.2139/ssrn.2983045
Hildebrandt M (2018) Algorithmic regulation and the rule of law. Philos Trans R Soc A 376(2128). https://doi.org/10.1098/rsta.2017.0355
Leenes RE (2011) Framing techno-regulation: an exploration of state and non-state regulation by technology. Legisprudence 5(2):143–160. https://doi.org/10.2139/ssrn.2182439
Madise Ü (2020) § 3 commentary to the Constitution of the Republic of Estonia. In: Madise Ü et al (eds) Eesti Vabariigi põhiseadus. Kommenteeritud väljaanne, 5th edn. Sihtasutus Iuridicum, Tartu
Martini M, Nink D (2018) Subsumtionsautomaten ante portas? - Zu den Grenzen der Automatisierung in (Rechtsbehelfs-)Verfahren. Deutsches Verwaltungsblatt (DVBl) 133(17):1128–1137
Martini M et al (2021) Staatshaftung für automatisierte Verwaltungsentscheidungen – Künstliche Intelligenz als Herausforderung für das Recht der staatlichen Ersatzleistungen. Verwaltungsarchiv 112:1–37
Merusk K et al (2012) § 3 commentary to the Constitution of the Republic of Estonia. In: Madise Ü et al (eds) Eesti Vabariigi põhiseadus. Kommenteeritud väljaanne. 3rd edn. Juura, Tallinn
Ministry of Justice of Estonia (2022) Haldusmenetluse seaduse muutmise ja sellega seonduvalt teiste seaduste muutmise seaduse eelnõu seletukiri (explanatory memorandum to the draft act amending the Administrative Procedure Act and other related Acts. file:///C:/Users/Kasutaja/Downloads/Seletuskiri%20(8).pdf
Narits R et al (2017) § 3 commentary to the constitution of the Republic of Estonia. In: Madise Ü et al (eds) Eesti Vabariigi põhiseadus. Kommenteeritud väljaanne, 4th edn. Juura, Tallinn
Parrest N (2009) Constitutional boundaries of transfer of public functions to [the] private sector in Estonia. Juridica Int 16:44–55. https://www.juridicainternational.eu/article_full.php?uri=2009_XVI_44_constitutional-boundaries-of-transfer-of-public-functions-to-private-sector-in-estonia
Pilving I, Mikiver M (2020a) A kratt as an administrative body: algorithmic decisions and principles of administrative law. Juridica Int 29:47–61. https://doi.org/10.12697/JI.2020.29.05
Pilving I, Mikiver M (2020b) Kratt haldusorganiks: algoritmilised otsused ja haldusõiguse põhimõtted. In: Parmas A (ed) Kohtute aastaraamat 2019. https://aastaraamat.riigikohus.ee/kratt-haldusorganiks-algoritmilised-otsused-ja-haldusoiguse-pohimotted/
Prell L (2020) § 35a. In: Bader J, Ronellenfitsch M (eds) Beck’scher Online-Kommentar VwVfG mit VwVG und VwZG. 49th edn. C.H. Beck, München
Stegmüller M (2018) Vollautomatische Verwaltungsakte – eine kritische Sicht auf die neuen § 24 I 3 und § 35a VwVfG. Neue Zeitschrift für Verwaltungsrecht (NVwZ), pp 353–358
Stein C, Drouin M (2017) Cognitive bias in the courtroom: combating the anchoring effect in criminal sentencing. https://ssrn.com/abstract=2991611
Stelkens P (2018) § 35a: Vollständig automatisierter Erlass eines Verwaltungsaktes. In: Stelkens P et al (eds) Verwaltungsverfahrensgesetz Kommentar. C.H. Beck, München
Tversky A, Kahneman D (1974) Judgment under uncertainty: heuristics and biases’. Science 185(4157):1124–1131. https://www.jstor.org/stable/1738360
Voßkuhle A, Kaufhold AK (2016) Grundwissen – Öffentliches Recht: Verwaltungsvorschriften. Juristische Schulung (JuS) 2016:314–316
Wills T (2019) Sweden: rogue algorithm stops welfare payments for up to 70,000 unemployed. Algorithm Watch, 25 February 2019. https://algorithmwatch.org/en/story/rogue-algorithm-in-sweden-stops-welfare-payments/
Ziekow J (2018) Das Verwaltungsverfahrensrecht in der Digitalisierung der Verwaltung. NVwZ, 1169–1172
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this chapter
Cite this chapter
Ebers, M., Trasberg, H. (2023). Delegation of Administrative Powers to AI Systems. In: Ebers, M., Tupay, P.K. (eds) Artificial Intelligence and Machine Learning Powered Public Service Delivery in Estonia. Data Science, Machine Intelligence, and Law, vol 2. Springer, Cham. https://doi.org/10.1007/978-3-031-19667-6_5
Download citation
DOI: https://doi.org/10.1007/978-3-031-19667-6_5
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-19666-9
Online ISBN: 978-3-031-19667-6
eBook Packages: Law and CriminologyLaw and Criminology (R0)