Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 8 of the 8 references it cites were found in a registry or an index; none of the 157 sources its claims cite is checked for support yet.

Slave to the Algorithm? Why a 'right to an explanation' is probably not the remedy you are looking for

Open access

Abstract

Cite as Lilian Edwards and Michael Veale, 'Slave to the Algorithm? Why a 'right to an explanation' is probably not the remedy you are looking for' (2017) 16 Duke Law and Technology Review 18–84. (First posted on SSRN 24 May 2017)Algorithms, particularly machine learning (ML) algorithms, are increasingly important to individuals’ lives, but have caused a range of concerns revolving mainly around unfairness, discrimination and opacity. Transparency in the form of a “right to an explanation” has emerged as a compellingly attractive remedy since it intuitively promises to “open the black box” to promote challenge, redress, and hopefully heightened accountability. Amidst the general furore over algorithmic bias we describe, any remedy in a storm has looked attractive.However, we argue that a right to an explanation in the EU General Data Protection Regulation (GDPR) is unlikely to present a complete remedy to algorithmic harms, particularly in some of the core “algorithmic war stories” that have shaped recent attitudes in this domain. Firstly, the law is restrictive, unclear, or even paradoxical concerning when any explanation-related right can be triggered. Secondly, even navigating this, the legal conception of explanations as “meaningful information about the logic of processing” may not be provided by the kind of ML “explanations” computer scientists have developed, partially in response. ML explanations are restricted both by the type of explanation sought, the dimensionality of the domain and the type of user seeking an explanation. However, “subject-centric" explanations (SCEs) focussing on particular regions of a model around a query show promise for interactive exploration, as do explanation systems based on learning a model from outside rather than taking it apart (pedagogical vs decompositional explanations ) in dodging developers' worries of IP or trade secrets disclosure.Based on our analysis, we fear that the search for a “right to an explanation” in the GDPR may be at best distracting, and at worst nurture a new kind of “transparency fallacy.” But all is not lost. We argue that other parts of the GDPR related (i) to the right to erasure ("right to be forgotten") and the right to data portability; and (ii) to privacy by design, Data Protection Impact Assessments and certification and privacy seals, may have the seeds we can use to make algorithms more responsible, explicable, and human-centred.

Overview

CoChat Summary

TL;DR

The paper argues that a GDPR “right to an explanation” is unlikely to provide a complete remedy for algorithmic harms because the legal triggers are restrictive or unclear and ML explanations may not supply legally “meaningful information about the logic of processing.” It suggests subject-centric and pedagogical explanations, plus other GDPR tools such as erasure, portability, DPIAs, privacy by design, and certification, as more promising avenues.

Key findings

  • The claimed GDPR right to explanation is legally shaky: Article 22 is limited to decisions based “solely” on automated processing that produce legal or similarly significant effects, with carve-outs and safeguards; Recital 71’s explanation mention is not in the main text, and recitals have uncertain binding force. In Sweeney’s AdSense study of 2,184 racially associated names, a black-identifying name was 25% more likely to get an arrest-related ad on one site, but it is hard to show a “decision” significantly affecting an individual.
  • ML explanations face barriers: model-centric explanations (MCEs) versus subject-centric explanations (SCEs), the “curse of dimensionality,” and tasks with many variables—e.g., LinkedIn claims over 100,000 variables per user—make meaningful explanation difficult. A system with 90% predictive accuracy would be expected to fail at least 10% of the time on new unseen data.
  • SCEs and pedagogical explanations are promising: SCEs focus on regions of a model around a query and can support interactive mental models; pedagogical/model-agnostic explanations query the model as an oracle and may avoid IP or trade-secret disclosure.
  • Other GDPR remedies may be more useful: erasure (Article 17) and portability (Article 20) have limits, especially for inferences; DPIAs (Article 35), privacy by design (Article 25), DPOs (Article 37), and certification/seals (Article 42) may help create better algorithmic governance. Fines can reach €20m or 4% of global turnover for severe breaches.

Methods

  • Design: interdisciplinary legal and technical conceptual analysis; no empirical experiment or dataset is used.
  • Data/sample: EU GDPR and Data Protection Directive provisions, EU case law, regulatory guidance, ML explanation literature, and media-reported “algorithmic war stories” such as Sweeney’s AdSense study, Google Spain, Target pregnancy profiling, and ProPublica’s criminal-risk assessment reporting.
  • Main analysis: textual/doctrinal reading of GDPR Articles 22, 15, 17, 20, 25, 35, 37, and 42; distinguishes MCEs/SCEs and decompositional/pedagogical explanations; assesses whether explanation rights and alternative remedies can meaningfully address algorithmic harms.

Limitations

  • Paper-stated: the broad scope across very different ML systems “may lead to some misleading generalisations”; there is a lack of empirical research on whether users seeking explanations are likely to be statistically anomalous “rare birds”; recitals’ legal status is murky and European case law is scarce.
  • Evident from the text: no empirical validation or user study is conducted; the analysis relies on selected war stories and literature; the legal interpretation is EU-focused and may be unsettled by future case law or regulatory guidance.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.31228/osf.io/97upg

Cites

8 of 8 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Subject-centred explanations in practice. The dotted line represents a machine learned decision boundary, where the ticks are classified one way, and the crosses another. This is a highly simplified, illustrative diagram that consciously omits uncertainty, or points misclassified during training in order to illustrate a broader point.
Figure 1 Subject-centred explanations in practice. The dotted line represents a machine learned decision boundary, where the ticks are classified one way, and the crosses another. This is a highly simplified, illustrative diagram that consciously omits uncertainty, or points misclassified during training in order to illustrate a broader point.From the PDF

Verification

Claims the paper makes

0 Valid · 26 Not valid · 150 Not checked · 173 Own results and interpretations

119 more claims

Create a free account to see all 349 claims and every source verdict.

See all claims — free account
  • Probably not a substantial claim. 193
  • It may state two separate claims as one. 79
  • The quote only partly states this claim. 1

Create a free account to see these claims.

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?

2 more claims about this paper

Create a free account to see every claim other papers make about this one.

See all claims — free account

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.

1 of 1

Citations

Cites

Show all 8 references

Cited by

See all 527

Metadata

Authors
Edwards, L., & Veale, M.
Affiliations
University of Strathclyde (GB)
Published
2017-11-18
OpenAlex
W4300071243
Cited by
524
References
8
Citations by year
2026: 26 · 2025: 32 · 2024: 52 · 2023: 130 · 2022: 57 · 2021: 78 · 2020: 89 · 2019: 37 · 2018: 15 · 2017: 8
Type
posted content
Language
en
Open access
Locations
  • Strathprints: The University of Strathclyde institutional repository (University of Strathclyde) (acceptedVersion)
  • Duke Law Scholarship Repository (Duke University) (submittedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • UCL Discovery (University College London) (submittedVersion)
  • OSF Preprints (OSF Preprints) (submittedVersion, cc-by)
  • LawArXiv (OSF Preprints) (submittedVersion, cc-by)
  • OSF Preprints (OSF Preprints) (submittedVersion)
Funders
  • Arts and Humanities Research Council
  • Engineering and Physical Sciences Research Council
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
  • Social Sciences › Law › Digitalization, Law, and Regulation
  • Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
Keywords
algorithmic transparency, algorithmic discrimination, algorithmic bias, General Data Protection Regulation, GDPR, right to be forgotten, data portability, privacy by design, data protection impact assessment, algorithmic accountability, algorithmic harms

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).

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: Slave to the Algorithm? Why a 'right to an explanation' is probably not the remedy you are looking for. https://app.cochat.ai/papers/10.31228/osf.io/97upg
The paper
Edwards, L., & Veale, M. (2017). Slave to the Algorithm? Why a 'right to an explanation' is probably not the remedy you are looking for. https://doi.org/10.31228/osf.io/97upg

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.