Pith. sign in

REVIEW 4 major objections 5 minor 34 references

Mapping Compliance: A Taxonomy for Political Content Analysis under the EU's Digital Electoral Framework

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper presents a nine-category codebook that maps the EU's DSA, TTPA, and electoral guidelines onto annotation categories for measuring political-content compliance.

desk verdict A transparent codebook for EU political-ad compliance whose central claim outruns its evidence: the legal mapping is careful, but several codes target platform-level obligations that a content sample cannot see, and the paper explicitly defers validation. read the letter →

arxiv 2501.01738 v1 pith:UBVZMWN4 submitted 2025-01-03 cs.CY

Marie-Therese Sekwenz , Rita Gsenger This is my paper · ORCID
classification cs.CY
keywords DigitalServicesActpoliticaladvertisingcontentanalysiscodebookelectoralintegritysystemicrisktransparencyEuropeanelections
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that the EU's new digital electoral rules — the Digital Services Act, the Transparency and Targeting of Political Advertising Regulation, and the Commission's electoral guidelines — can be rendered as a concrete content-analysis taxonomy. It proposes a codebook with nine new annotation categories (Ro-1 to Ro-9) covering gender-based violence, sponsor identification, targeting disclosure, ad repositories, ad labelling, influencer content, synthetic media, silence periods, and fact-checking, each tied to specific legal provisions. If the taxonomy is right, external auditors and platform risk teams could sample user-generated and paid political content and systematically score compliance with electoral-integrity obligations. The paper does not yet test the codebook; its stated next step is empirical validation.

What carries the argument

The load-bearing mechanism is the legal-doctrinal mapping from each annotation code to named provisions in the three instruments. The codebook's unit is a code: a short alphanumeric label such as Ro-5 for political ad labelling, with a source line citing provisions like Articles 26 and 39 DSA or Article 11 TTPA, a qualification that determines what counts, a definition, and an example to anchor coder judgment. The nine new regulatory codes are designed to capture the transparency and risk-mitigation duties that older electoral and disinformation taxonomies did not cover. The reliability machinery is inter-coder agreement measured by Krippendorff's Alpha on sample content drawn by cluster or random sampling.

What would settle it

A direct test would be to take a sample of political ads that regulators have already found non-compliant, such as ads with missing sponsor information or synthetic media without labels, have independent coders apply the Ro-1 to Ro-9 codes, and check whether the codebook flags all known violations without excessive false positives. A second check is legal: if a code's source provisions were mis-cited or a mandatory duty such as the Article 39 DSA repository fields is not represented, the mapping fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that a legally grounded codebook can operationalize the compliance duties in the DSA, TTPA, and G-E-DSA for empirical content analysis. By doing doctrinal close reading of the three instruments and merging them with existing disinformation and election-risk category sets, the authors construct a three-part annotation system: electoral-rights codes, regulatory codes Ro-1 through Ro-9, and disinformation codes. Each code carries a source provision, a qualification, a definition, and examples, so that annotators can apply it to text, image, video, and audio content. The intended use is to test systemic risks under Article 34(1)(c) DSA, specifically negative effects on civic discourse and electoral processes, and to support external audit of platform claims under Article 37 DSA. The authors are explicit that the taxonomy is a tool for future validation, not a validated instrument.

Load-bearing premise

The load-bearing premise is that the paper's reading of the DSA, TTPA, and G-E-DSA is legally correct and complete enough that applying the nine codes to content actually measures compliance.

Editorial extensions

If this is right

  • Auditors and civil-society monitors can apply the codebook to a sample of ads and user posts to produce a structured compliance report across the three regulatory domains.
  • The taxonomy gives Article 34(1)(c) DSA a measurable empirical meaning: negative effects on civic discourse and electoral processes become observable annotation categories rather than an open-ended legal phrase.
  • Platforms can use the same categories when building ad repositories, labeling synthetic media, enforcing silence periods, and responding to fact-checker labels, aligning content moderation with the G-E-DSA guidance.
  • Because each code cites specific provisions, the framework can support automated classifiers or language-model detection pipelines that need a labeled dataset aligned with EU law.
  • Comparative studies across member states and elections become possible using a shared codebook keyed to the same legal texts.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the paper's legal mapping is untested; the most decisive next study would be an inter-coder reliability test on real ad-repository data from a very large online platform, with the codes applied by coders who have no stake in the taxonomy.
  • Editorial inference: the taxonomy may need extension for non-ad political content and for organic influencer posts whose status under the TTPA definition of political advertising is unsettled; the paper itself notes the definitional complexity of political content.
  • Editorial inference: if validated, the codebook could become a bridge between the DSA audit regime and the AI Act's transparency duties for synthetic content, since it already codes deepfakes and AI-generated political media.
  • Editorial inference: the absence of a test means the taxonomy should be read as a hypothesis about which legal provisions are observable in content, not as evidence of what platforms currently fail to do.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. This paper proposes a taxonomy/codebook for content analysis of political content and political ads under the EU's Digital Services Act (DSA), the Regulation on Transparency and Targeting of Political Advertising (TTPA), and the Commission's Guidelines for electoral processes (G-E-DSA). It introduces nine new annotation categories (Ro-1 to Ro-9), maps each to legal provisions, and augments them with previously published categories for electoral-rights risks and disinformation. The authors argue that the resulting taxonomy enables systematic content analysis of user-generated and ad content to assess compliance with regulatory mandates and to support systemic risk assessments under Art. 34 and Art. 37 DSA. The paper contains a legal doctrinal review, a table of the nine new categories, and an annex codebook, but no empirical validation of the codebook and no specification of how platform-level codes can be applied to individual content items.

Significance. If validated, this taxonomy would provide a useful bridge between EU regulatory texts and empirical content analysis, especially for external audits under Art. 37 DSA and for research on Art. 34(1)(c) DSA systemic electoral risks. The legal mapping is transparent and the codebook structure (code, source, definition, example) is a sensible, replicable format. The paper is best understood as a well-organized doctrinal proposal rather than a completed empirical instrument; its main value at this stage is as a starting point for annotation studies, provided the proposed categories are pretested and platform-level data are integrated.

major comments (4)
  1. [§4 and Annex 7.2 (Ro-4, Ro-8, Ro-9)] The central claim that the taxonomy 'enables systematic content analysis ... to assess compliance' is not supported by the proposed annotation procedure because several codes apply to platform-level obligations rather than to individual content items. Ro-4 ('Ad Repository Requirements') requires a judgment about whether a platform maintains a public repository, Ro-8 ('Silence Period Enforcement') requires knowing whether an ad ran during a national silence period, and Ro-9 ('Monitoring and Fact-checking') requires knowing whether the platform attached a fact-check label; Ro-2, Ro-3, and Ro-5 similarly depend on ad-repository metadata or platform display state. Section 4 describes coders applying each annotation category to a sample of content, but no separate instrument or data linkage is specified for these platform-level codes, so the codebook as presented cannot deliver compliance measurement on individual content items without an additional platform-audit layer.
  2. [§4 Methodology and §6 Conclusion] The claim that the taxonomy enables systematic content analysis is not empirically tested. The methodology states that coding consistency is measured through Krippendorff's Alpha and pre-tests, but the paper reports no pretest results, no inter-coder reliability coefficient, and no annotated sample; the conclusion explicitly defers validation to 'future work.' As it stands, the contribution is a proposed codebook, not a validated instrument, and the abstract's wording should be revised accordingly or supplemented with a pilot annotation study.
  3. [Annex 7.2, Ro-8] The legal mapping for Ro-8 ('Silence Period Enforcement') is incomplete: the source column cites only 'Rec. 14,' while the body text in Section 3.3 says platforms are 'encouraged to respect' silence periods, reflecting that silence periods are primarily imposed by Member State law rather than by a directly binding EU obligation. Since the paper's central claim rests on the correctness and completeness of the code-to-provision mapping, this category needs a fuller legal basis or a re-framing as a Member-State-dependent indicator.
  4. [Table 1 and §5] Ro-1 is tied to 'Rec 81 DSA, Art 34(1)(d) DSA' and to the Directive on combating violence against women, whereas the paper's stated focus is the electoral-process risk under Art. 34(1)(c) DSA; the relationship between gender-based violence and electoral-process risk is asserted rather than argued, so the inclusion of this category within an electoral-risk codebook needs a clearer justification.
minor comments (5)
  1. [§4] There is a typo in the definition of 'qualification': it reads 'e.g.,hoxes' and should be 'e.g., hoaxes.'
  2. [Throughout] The abbreviation 'TTP' is used interchangeably with 'TTPA' (e.g., RQ2 and Section 3.2); please use a single abbreviation consistently.
  3. [Annex 7.2, Ro-1 source list] The source list mixes article ranges and individual articles in a way that is ambiguous ('Art. 3-9 Directive ... Art. 3 Directive ... Art. 4 Forced marriage'); please format each source as a complete, unambiguous citation.
  4. [§4, reference to Krippendorff's Alpha] The cited reference (Zapf et al.) is a comparison of inter-rater coefficients, not the primary source for Krippendorff's Alpha; please cite Krippendorff's 'Content Analysis' directly or a methodology text that defines the coefficient.
  5. [§5] The text states that the taxonomy includes 'misrepresentation in political advertisements' as a category, but no such standalone category appears in Table 1 or the annex; please align the narrative with the actual codebook.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the taxonomy is a legal-doctrinal mapping with no fitted parameters, predictions, or self-citation load-bearing.

full rationale

The paper's central contribution is a codebook of annotation categories derived directly from cited legal provisions (DSA, TTPA, G-E-DSA). No equations or fitted quantities are present, so no 'prediction' can reduce to an input by construction. The new Ro-1 to Ro-9 categories are each explicitly tied to legal articles (e.g., Ro-4 to Art. 13 TTPA and Art. 39 DSA; Ro-7 to G-E-DSA Sec. 3.2.3), making the mapping independently checkable against primary law. Prior work by the authors (Kübler et al., Wagner et al.) is cited as the source of older categories and coding guidance, but the new categories do not depend on those citations for their justification; the legal texts are the load-bearing evidence. The paper expressly disclaims empirical validation ('Future work will focus on ... conducting empirical test to validate and operationalize the taxonomy'), so there is no overclaim that could hide a fitted-input-as-prediction. Any concerns about unit-of-analysis mismatch between content-level codes and platform-level obligations are correctness risks, not circularity. The derivation chain is self-contained in the sense that the taxonomy is a transparent operationalization of the regulations it names.

Assumptions & free parameters 0 free parameters · 4 assumptions · 1 invented entities

No quantitative free parameters are involved. The taxonomy depends on the correctness and completeness of legal interpretation, transferability of prior codebooks, and future empirical validation, none of which is demonstrated in this paper.

assumptions (4)
  • domain assumption The selected legal sources (DSA, TTPA, G-E-DSA) sufficiently cover the EU digital electoral regulatory landscape.
    Stated in Section 1: 'To our knowledge, the selection of these legal sources adequately portrays the relevant regulatory landscape.' If other rules (e.g., AI Act or national laws) are required for compliance, the taxonomy is incomplete.
  • domain assumption The legal-doctrinal interpretation of each cited article is correct and maps one-to-one to the annotation category.
    Each Ro-category cites articles; the paper does not provide separate legal validation or expert review.
  • domain assumption Prior frameworks by Kübler et al. and Kapantai et al. are valid and transferable to the new regulatory context.
    The electoral-rights and disinformation codes in the annex are adopted from these sources; the paper assumes they remain applicable.
  • domain assumption Content analysis with these codes can be reliably applied by coders and yields meaningful compliance assessments.
    The paper proposes Krippendorff's alpha but reports no reliability test; future work is said to validate the taxonomy.
invented entities (1)
  • Ro-1 to Ro-9 annotation categories
    purpose: Operationalize DSA, TTPA, and G-E-DSA obligations into content-analysis codes for political ads and user posts.
    The nine categories are introduced in Section 5 and Table 1 with definitions and examples, but no empirical inter-coder reliability, pilot dataset, or external validation is provided. They are derived from legal texts and prior frameworks rather than tested against data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Mapping Compliance: A Taxonomy for Political Content Analysis under the EU's Digital Electoral Framework." pith.science (2026). https://pith.science/paper/UBVZMWN4

@misc{pith2026250101738,
  author       = {Pith},
  title        = {Pith review of: Mapping Compliance: A Taxonomy for Political Content Analysis under the EU's Digital Electoral Framework},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UBVZMWN4}},
  note         = {Machine review of arXiv:2501.01738}
}
read the original abstract

The rise of digital platforms has transformed political campaigning, introducing complex regulatory challenges. This paper presents a comprehensive taxonomy for analyzing political content in the EU's digital electoral landscape, aligning with the requirements set forth in new regulations, such as the Digital Services Act. Using a legal doctrinal methodology, we construct a detailed codebook that enables systematic content analysis across user-generated and political ad content to assess compliance with regulatory mandates.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 33 canonical work pages

  1. [1]

    Introduction After the 2016 US elections and the Brexit referendum, political consensus formed in the EU that novel rules for the context of elections are needed to increase online protection.1 The digital dimensions of political debate and their potential negative consequences can have manifestations offline, which was illustrated by the storm on the US ...

    work page 2016

  2. [2]

    Political Advertising and Regulation Political campaigns and the creation of political ads play an important role in elections. Politicians create comparatively cheap advertising strategies (in contrast to TV advertising) and fine-tune their message, visual, or appearance to target a specific voter audience, e.g., through gender filtering.12 This separati...

    work page 2018

  3. [6]

    threshold of illegality.7 Following the fine line between illegality and harmful speech, however, is posing questions to the protection of fundamental rights, such as freedom of speech or the right to free and fair elections.8 Furthermore, if concepts like mis- or disinformation are not tangible and clearly definable,9 the automatic detection of such viol...

    work page 2021

  4. [7]

    Martin Sichert: Billionaire George Soros is known to want to make the world

    Annex: Codebook 7.1 Electoral Rights Categories43 Codename & Source Definition Example C-1: Candidates - Right to stand for elections Actors interested in restricting electoral competition to ensure electoral victory (of the government) might employ false claims that eligibility conditions are not given in order to exclude certain candidates. Fabricated c...

    work page 2021

  5. [14]

    19 FLYVERBOM, Transparency: Mediation and the management of visibilities, IJoC 10, 2016, p

    work page 2016

  6. [16]

    However, it is very complex to define ’political content’ on platforms (for example, what constitutes political advertising in cases of influencer content21 and how to label or filter it automatically. Several platforms, such as TikTok, have claimed not to allow political content and failed.22 How can such claims be meaningfully and empirically tested in ...

    work page 2024

  7. [17]

    https://www.washingtonpost.com/politics/2024/10/17/tiktok-political-ads-ban-report-global-witness/ (accessed on 12 November 2024),

    22 OREMUS, TikTok fails a test of its ban on political ads, Washington Post. https://www.washingtonpost.com/politics/2024/10/17/tiktok-political-ads-ban-report-global-witness/ (accessed on 12 November 2024),

    work page 2024

  8. [18]

    https://digital-strategy.ec.europa.eu/en/news/commission-launches-new-database-track-digital-services-terms-and-conditions (accessed on 14 November, 2024),

    23 EUROPEAN COMMISSION, Commission launches new database to track digital services terms and conditions. https://digital-strategy.ec.europa.eu/en/news/commission-launches-new-database-track-digital-services-terms-and-conditions (accessed on 14 November, 2024),

    work page 2024

Show all 34 references
  1. [20]

    Images to Falsely Suggest Taylor Swift Endorsed Him, The New York Times

    26 VIGDOR, Trump Promotes A.I. Images to Falsely Suggest Taylor Swift Endorsed Him, The New York Times. https://www.nytimes.com/2024/08/19/us/politics/trump-taylor-swift-ai-images.html (accessed on 12 September 2024),

    2024

  2. [21]

    https://www.theguardian.com/technology/article/2024/aug/22/fake-biden-robocalls-fine-lingo-telecom (accessed on 12 September 2024),

    27 ASSOCIATED PRESS, Company that sent fake Biden robocalls in New Hampshire agrees to $1m fine, The Guardian. https://www.theguardian.com/technology/article/2024/aug/22/fake-biden-robocalls-fine-lingo-telecom (accessed on 12 September 2024),

    2024

  3. [22]

    28 MEAKER, Slovakia’s Election Deepfakes Show AI Is a Danger to Democracy. Wired. https://www.wired.com/story/slovakias-election-deepfakes-show-ai-is-a-danger-to-democracy/ (accessed on 02 January 2025),

    2025

  4. [23]

    29 KÜBLER ET AL, The 2021 German Federal Election on Social Media: Analysing Electoral Risks Created by Twitter and Facebook. Proc. of the 56th Hawaii International Conference on System Sciences (HICSS-56), Maui

    2021

  5. [25]

    14 JINGNAN, AI-generated images have become a new form of propaganda this election season, NPR, https://www.opb.org/article/2024/10/18/ai-generated-propaganda-election-hurricane/ (accessed on 11 November 2024),

    2024

  6. [26]

    32 Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regul...

    2016

  7. [27]

    https://digital-strategy.ec.europa.eu/en/library/guidelines-providers-vlops-and-vloses-mitigation-systemic-risks-electoral-processes (accessed on 19 November 2024),

    35 EUROPEAN COMMISSION, Commission on Guidelines for providers of Very Large Online Platforms and Very Large Online Search Engines on the mitigation of systemic risks for electoral processes pursuant to the Digital Services Act. https://digital-strategy.ec.europa.eu/en/library...

    2024

  8. [28]

    37 KÜBLER ET AL, The 2021 German Federal Election on Social Media: Analysing Electoral Risks Created by Twitter and Facebook. Proc. of the 56th Hawaii International Conference on System Sciences (HICSS-56), Maui

    2021

  9. [29]

    38 KAPANTAI ET AL, A systematic literature review on disinformation: Toward a unified taxonomical framework, New Media & Society 23, 5, 2021, p

    2021

  10. [30]

    34 DSA and testing systemic risk assessment reports as external independent auditors on a sample basis, according to Art

    We propose content analysis as a qualitative methodology for conducting systemic risk assessments under the DSA in line with Art. 34 DSA and testing systemic risk assessment reports as external independent auditors on a sample basis, according to Art. 37 DSA and the Delegated ...

    2022

  11. [31]

    Conclusion This paper introduces a comprehensive taxonomy for political content and advertising under the EU’s digital electoral framework, encompassing the DSA, TTPA, and G-E–DSA. By developing a structured approach for annotating, reporting, and monitoring political content,...

    2024

  12. [33]

    Attila Hildmann: Bearbock verurteilte angriffe der hamas auf israel: 'Wir stehen an der Seite der Israelis' […] Bearbock stands on the side of Israel because Bearbock is Jewish

    "Attila Hildmann: Bearbock verurteilte angriffe der hamas auf israel: 'Wir stehen an der Seite der Israelis' […] Bearbock stands on the side of Israel because Bearbock is Jewish" E-4: Candidates - Election polls Actors interested in (de-)legitimizing the elections or harming/p...

    2020

  13. [34]

    A political ad shown during the silence period prior to voting day

    Platforms must respect national silence periods (ad bans before elections) to avoid undue influence on voters. A political ad shown during the silence period prior to voting day. Ro-9: Monitoring and Fact-checking (G-E–DSA Sec. 3.3) Platforms must label content identified by f...

    2021

  14. [40]

    This provision demands the identifiable and consistent indication that the content is labeled as advertising content (Art

    Regarding advertising, the DSA furthermore defines minimal general rules for advertising online in Art 26 DSA. This provision demands the identifiable and consistent indication that the content is labeled as advertising content (Art. 26 (1) (a) DSA), for whom the advertising i...

    2024

  15. [41]

    Federal Elections, 2004-2020, Journal of Quantitative Description: Digital Media 2,

    17 SHEINGATE/SCHARF/DELAHANTY, Digital Advertising in U.S. Federal Elections, 2004-2020, Journal of Quantitative Description: Digital Media 2,

    2004

  16. [110]

    20 WHO TARGETS ME, A Goldilocks Zone for political ads https://whotargets.me/en/a-goldilocks-zone-for-political-ads/ (accessed on 18 November 2024),

    2024

  17. [119]

    Recital 95 DSA additionally stresses the need for information about targeting criteria and the relationship to vulnerable groups and situations

    concerning health or data concerning a natural person’s sex life or sexual orientation”. Recital 95 DSA additionally stresses the need for information about targeting criteria and the relationship to vulnerable groups and situations. A demand echoed in the Delegated Regulation...

    2022

  18. [121]

    2 NG/CRUICKSHANK/CARLEY, Cross-platform information spread during the January 6th capitol riots, Social Network Analysis and Mining 12, p. 133; ASSOCIATED PRESS, „Anti-corona“ extremists try to storm German parliament, The Guardian, https://www.theguardian.com/world/2020/aug/2...

    2020

  19. [277]

    4 Regulation (EU) 2024/900 of the European Parliament and of the Council of 13 March 2024 on the transparency and targeting of political advertising (Text with EEA relevance). – The Commission’s Guidelines on Mitigation of Systemic Risks for Electoral Processes under the DSA (...

    2024

  20. [326]

    9 KAPANTAI ET AL, A systematic literature review on disinformation: Toward a unified taxonomical framework, New Media & Society 23, 5, 2021, p

    2021

  21. [1301]

    11 KÜBLER ET AL, The 2021 German Federal Election on Social Media: Analysing Electoral Risks Created by Twitter and Facebook

    10 BUCHANAN ET AL, Truth, Lies, and Automation: How Language Models Could Change Disinformation; VYKOPAL ET AL, Disinformation Capabilities of Large Language Models, ArXiv abs/2311.08838. 11 KÜBLER ET AL, The 2021 German Federal Election on Social Media: Analysing Electoral Ri...

    2021 arXiv

  22. [2020]

    3 Regulation (EU) 2022/2065 of the European Parliament and of the Council of 19 October 2022 on a Single Market For Digital Services and amending Directive 2000/31/EC (Digital Services Act) (Text with EEA relevance) OJ L

    2022

  23. [2021]

    16 PACHILAKIS ET AL, YourAdvalue: Measuring Advertising Price Dynamics without Bankrupting User Privacy, Proceedings of the 2022 ACM SIGMETRICS/IFIP PERFORMANCE Joint International Conference on Measurement and Modeling of Computer Systems 2022, p

    2022

  24. [2022]

    18 FACEBOOK HELP CENTER, Why am I seeing ads from an advertiser on Facebook? https://www.facebook.com/help/794535777607370/ (accessed on 31 December 2024),

    2024

  25. [2023]

    24 Delegated Regulation supplementing Regulation (EU) 2022/2065 of the European Parliament and of the Council, by laying down rules on the performance of audits for very large online platforms and very large online search engines, 20.10.2023. 25 EUROPEAN COMMISSION, Commission...

    2022

  26. [2024]

    6 EUROPEAN COMMISSION, Commission opens formal proceedings against X under the DSA https://ec.europa.eu/commission/presscorner/detail/en/ip_23_6709 (accessed on 19 November 2024)), 2023; EUROPEAN COMMISSION, Commission opens formal proceedings against TikTok https://ec.europa....

    2024

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.