pith. sign in

arxiv: 2606.02347 · v1 · pith:4Y7CUUNFnew · submitted 2026-06-01 · 💻 cs.CY

Are Algorithm Registers Transparent? Perspectives from Germany

Iman Peljto , Xenia Heilmann , Mattia Cerrato This is my paper

Pith reviewed 2026-06-28 12:15 UTC · model grok-4.3

classification 💻 cs.CY
keywords algorithm registerstransparencyGermanyAI governancepublic administrationauditMaKILernende Systeme
0
0 comments X

The pith

German algorithm registers like MaKI and Lernende Systeme require adaptations to meet proposed national transparency goals.

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

The paper extracts structured checklists from the transparency goals and subgoals in the Alina Lorenz 2025 proposal for a German national AI register. It applies these checklists to audit the two main existing German initiatives, MaKI and Lernende Systeme. The audit identifies multiple gaps, leading to the conclusion that adaptations are needed for the registers to function as useful transparency instruments. This evaluation addresses the fragmented landscape of algorithm information in German public administration and provides publicly available checklists for future audits or designs.

Core claim

By repurposing the Lorenz proposal as an audit instrument and applying the derived checklists, the authors determine that MaKI and Lernende Systeme fulfill only a subset of the outlined transparency goals and subgoals. Several adaptations are therefore likely required for these registers to serve as useful transparency instruments, with additional proposals for a visualization of transparency levels and concrete action items to improve the platforms.

What carries the argument

Structured checklists extracted from the transparency goals and subgoals of the Alina Lorenz (2025) proposal, translated into English and used to evaluate existing algorithm registers.

If this is right

  • The existing registers should incorporate additional categories of information on algorithms to align with the proposed goals.
  • A visualization method can be used to map and communicate the transparency levels achieved by different registers.
  • Specific action items can direct improvements to MaKI, Lernende Systeme, and similar platforms.
  • The checklists can support the design of new registers or audits of other initiatives in Germany and beyond.

Where Pith is reading between the lines

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

  • The same checklist approach could be applied to algorithm registers in other European countries to identify comparable gaps.
  • Closing the identified gaps might improve public access to information on government use of algorithms.
  • The checklists offer a ready tool for practitioners to benchmark new or existing registers against a consistent set of transparency criteria.

Load-bearing premise

The transparency goals and subgoals formulated in the Alina Lorenz proposal provide an appropriate and complete benchmark for whether existing algorithm registers deliver meaningful transparency.

What would settle it

An independent application of the same checklists to MaKI and Lernende Systeme that finds they already satisfy most or all listed goals and subgoals would indicate that the registers meet the transparency benchmark without needing adaptations.

Figures

Figures reproduced from arXiv: 2606.02347 by Iman Peljto, Mattia Cerrato, Xenia Heilmann.

Figure 1
Figure 1. Figure 1: Methodology overview. clearly justified by privacy, security, or safety concerns. We proceed in five steps ( [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
read the original abstract

Algorithm registers are public-facing databases that display basic information about algorithms employed in public administration. While several such registers exist across Europe and globally, their capacity to deliver meaningful transparency remains contested. In Germany, the landscape is notably fragmented: no federal-level register exists, yet at least five state- and federal-level initiatives publish information about AI systems with varying scopes and objectives. A recent conceptual proposal by Alina Lorenz (2025), outlines technical and governance requirements for a national AI transparency register in Germany. We repurpose this proposal as an audit instrument, extracting structured checklists from the transparency goals and subgoals it formulates. The resulting checklists, translated from German into English, is made publicly available to support practitioners auditing existing registers or designing new ones. We apply this framework to conduct an external audit of the two main existing German transparency initiatives, MaKI and Lernende Systeme, evaluating the extent to which they fulfill the proposed goals. Our audit reveals that several adaptations are likely needed for these registers to serve as an useful transparency instrument. We further propose a visualization of register transparency levels and derive concrete action items for improving existing German platforms.

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.

Referee Report

2 major / 2 minor

Summary. The manuscript repurposes the transparency goals and subgoals from Lorenz (2025) into structured checklists, applies them in an external audit of the MaKI and Lernende Systeme registers, concludes that several adaptations are needed for these to serve as useful transparency instruments, makes the translated checklists publicly available, and proposes a visualization of transparency levels plus concrete action items.

Significance. If the audit results hold, the work supplies a reusable, publicly available checklist instrument for evaluating algorithm registers and identifies concrete gaps in the two main German initiatives. The decision to release the checklists supports reproducibility and direct use by practitioners designing or auditing similar platforms.

major comments (2)
  1. [Abstract] Abstract: the central claim that 'several adaptations are likely needed' is derived solely from scoring against the Lorenz (2025) goals and subgoals, yet the manuscript supplies no comparative mapping to other established frameworks (EU AI Act Articles 13/86, OECD AI Principles, or German administrative transparency statutes) and no derivation showing why these goals are complete. This assumption is load-bearing for the audit conclusions.
  2. [Abstract] Abstract (audit description): the evaluation steps are described only at a high level with no information on scoring criteria, inter-rater reliability, or how subgoals were operationalized into checklist items. Without these details the reported shortfalls cannot be independently verified.
minor comments (2)
  1. [Abstract] Abstract: 'an useful transparency instrument' is grammatically incorrect and should read 'a useful transparency instrument'.
  2. [Abstract] Abstract: subject-verb agreement error in 'The resulting checklists, translated from German into English, is made publicly available' (plural subject requires 'are').

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and will revise the manuscript to improve clarity and provide additional context where appropriate, without changing the core scope or findings of the audit.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that 'several adaptations are likely needed' is derived solely from scoring against the Lorenz (2025) goals and subgoals, yet the manuscript supplies no comparative mapping to other established frameworks (EU AI Act Articles 13/86, OECD AI Principles, or German administrative transparency statutes) and no derivation showing why these goals are complete. This assumption is load-bearing for the audit conclusions.

    Authors: The manuscript's explicit scope is to repurpose and apply the Lorenz (2025) proposal as a context-specific audit instrument for German registers, not to establish it as a complete or universal framework. We selected it due to its recency and direct relevance to the German national register debate. We agree that situating it against other frameworks would strengthen the paper. We will add a concise paragraph in the Introduction that maps key Lorenz subgoals to EU AI Act Articles 13/86, OECD AI Principles, and relevant German administrative transparency requirements, while stating the rationale for the chosen scope. This revision provides context but does not alter the audit results or conclusions. revision: yes

  2. Referee: [Abstract] Abstract (audit description): the evaluation steps are described only at a high level with no information on scoring criteria, inter-rater reliability, or how subgoals were operationalized into checklist items. Without these details the reported shortfalls cannot be independently verified.

    Authors: The abstract is intentionally concise. The full manuscript's Methods section describes the translation of Lorenz subgoals into checklist items and the application process. However, we acknowledge that explicit details on the scoring rubric, operationalization criteria, and inter-rater procedure (two authors scoring independently with consensus resolution) are not sufficiently prominent. We will expand the Methods section with a clear description of the operationalization approach, scoring criteria, and reliability assessment, add a brief methods summary sentence to the abstract, and include an appendix with example mappings. The already-public checklists support external verification. revision: yes

Circularity Check

0 steps flagged

No significant circularity; external benchmark used

full rationale

The paper extracts checklists from the transparency goals in Alina Lorenz (2025), an independent proposal by a non-author, then applies them to audit MaKI and Lernende Systeme. This comparison does not reduce by construction to the present authors' inputs or prior definitions. No self-citation load-bearing steps, no self-definitional relations, and no fitted predictions appear in the derivation chain. The central claim follows from applying an external standard rather than renaming or fitting the authors' own material.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on treating the Lorenz proposal as the authoritative standard for transparency without additional empirical validation of its goals within this paper.

axioms (1)
  • domain assumption Transparency goals stated in the Lorenz proposal can be reliably translated into structured, auditable checklists.
    The paper states that it extracts structured checklists from the transparency goals and subgoals formulated in the proposal.

pith-pipeline@v0.9.1-grok · 5727 in / 1169 out tokens · 30359 ms · 2026-06-28T12:15:14.350252+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

26 extracted references · 14 canonical work pages · 1 internal anchor

  1. [1]

    2025 , note =

    Marktplatz der KI-M. 2025 , note =

    2025

  2. [2]

    2025 , note =

    Alina Lorenz , title =. 2025 , note =

    2025

  3. [3]

    2024 , volume =

    Regulation (EU) 2024/1689 (Artificial Intelligence Act) , journal =. 2024 , volume =

    2024

  4. [4]

    2025 , file =

    Making. 2025 , file =

    2025

  5. [5]

    Philosophy & Technology , author =

    Artificial. Philosophy & Technology , author =. 2020 , pages =. doi:10.1007/s13347-020-00434-3 , language =

    work page doi:10.1007/s13347-020-00434-3 2020

  6. [6]

    Techné Research in Philosophy and Technology , author =

    Dutch. Techné Research in Philosophy and Technology , author =. 2022 , pages =. doi:10.5840/techne202323172 , number =

    work page doi:10.5840/techne202323172 2022

  7. [7]

    Information Polity , author =

    Algorithm. Information Polity , author =. doi:10.1177/15701255241297107 , number =

    work page doi:10.1177/15701255241297107

  8. [8]

    Transparent to whom?

    Kemper, Jakko and Kolkman, Daan , month = dec, year =. Transparent to whom?. Information, Communication & Society , publisher =. doi:10.1080/1369118X.2018.1477967 , abstract =

    work page doi:10.1080/1369118x.2018.1477967 2018

  9. [9]

    Administrative Law Review , author =

    Transparency and. Administrative Law Review , author =. 2019 , pages =

    2019

  10. [10]

    Land Cover Mapping Using Ensemble Feature Selection Methods

    Algorithmic. Digital Journalism , author =. doi:10.1080/21670811.2016.1208053 , number =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1080/21670811.2016.1208053 2016

  11. [11]

    and Selman, C

    Leslie, M. and Selman, C. , year =. Securing meaningful transparency of public sector use of

  12. [12]

    State of the

    Valderrama, Juan and. State of the

  13. [13]

    Accountable Artificial Intelligence: Holding Algorithms to Account , shorttitle =

    Accountable. Public Administration Review , author =. 2021 , note =. doi:10.1111/puar.13293 , abstract =

    work page doi:10.1111/puar.13293 2021

  14. [14]

    Explaining

    Grimmelikhuijsen, Stephan , year =. Explaining. doi:10.1111/puar.13483 , journal =

    work page doi:10.1111/puar.13483

  15. [15]

    , author Meijer, A

    Legitimacy of. Perspectives on Public Management and Governance , author =. 2022 , pages =. doi:10.1093/ppmgov/gvac008 , number =

    work page doi:10.1093/ppmgov/gvac008 2022

  16. [16]

    Introduction to special issue algorithmic transparency in government:

    Giest, Sarah and Grimmelikhuijsen, Stephan , year =. Introduction to special issue algorithmic transparency in government:. doi:10.3233/IP-200010 , journal =

    work page doi:10.3233/ip-200010

  17. [17]

    Haataja, Meeri and van de Fliert, Linda and Rautio, Pasi , month = sep, year =. Public

  18. [18]

    Mitchell, Margaret and Wu, Simone and Zaldivar, Andrew and Barnes, Parker and Vasserman, Lucy and Hutchinson, Ben and Spitzer, Elena and Raji, Inioluwa Deborah and Gebru, Timnit , year =. Model. doi:10.1145/3287560.3287596 , booktitle =

    work page doi:10.1145/3287560.3287596

  19. [19]

    European Law Open , author =

    Reclaiming transparency: contesting the logics of secrecy within the. European Law Open , author =. 2023 , keywords =. doi:10.1017/elo.2022.47 , abstract =

    work page doi:10.1017/elo.2022.47 2023

  20. [20]

    2025 , annote =

    Guidelines on the definition of an artificial intelligence system established by. 2025 , annote =

    2025

  21. [21]

    Think about the stakeholders first!

    Bell, Andrew and Nov, Oded and Stoyanovich, Julia , year =. Think about the stakeholders first!. doi:10.1017/dap.2023.8 , journal =

    work page doi:10.1017/dap.2023.8 2023

  22. [22]

    ECIS 2024 Proceedings , author =

    Defining and. ECIS 2024 Proceedings , author =. 2024 , file =

    2024

  23. [23]

    Algorithm registration in the

  24. [24]

    Seeing without Knowing:

    Ananny, Mike and Crawford, Kate , month = mar, year =. Seeing without knowing:. New Media & Society , publisher =. doi:10.1177/1461444816676645 , abstract =

    work page doi:10.1177/1461444816676645

  25. [25]

    Datasheets for datasets , volume =. Commun. ACM , author =. 2021 , pages =. doi:10.1145/3458723 , abstract =

    work page doi:10.1145/3458723 2021

  26. [26]

    Beyond the "

    Murad, Maya , month = jun, year =. Beyond the "