Algorithmic profiling as a source of hermeneutical injustice
Open accessThe DOI matches its Crossref record; 42 of the 42 references it cites were found in a registry or an index; 7 of the 25 sources its claims cite were checked for support.
Abstract
It is well-established that algorithms can be instruments of injustice. It is less frequently discussed, however, how current modes of AI deployment often make the very discovery of injustice difficult, if not impossible. In this article, we focus on the effects of algorithmic profiling on epistemic agency. We show how algorithmic profiling can give rise to epistemic injustice through the depletion of epistemic resources that are needed to interpret and evaluate certain experiences. By doing so, we not only demonstrate how the philosophical conceptual framework of epistemic injustice can help pinpoint potential, systematic harms from algorithmic profiling, but we also identify a novel source of hermeneutical injustice that to date has received little attention in the relevant literature, what we call epistemic fragmentation. As we detail in this paper, epistemic fragmentation is a structural characteristic of algorithmically-mediated environments that isolate individuals, making it more difficult to develop, uptake and apply new epistemic resources, thus making it more difficult to identify and conceptualise emerging harms in these environments. We thus trace the occurrence of hermeneutical injustice back to the fragmentation of the epistemic experiences of individuals, who are left more vulnerable by the inability to share, compare and learn from shared experiences.
Overview
CoChat Summary
TL;DR
The paper argues that algorithmic profiling can produce hermeneutical injustice by depleting the epistemic resources people need to interpret their experiences. It introduces epistemic fragmentation—a structural feature of algorithmically mediated environments that isolates individuals—as a novel source of such injustice, impairing the formation, uptake, and application of epistemic resources.
Key findings
- Algorithmic profiling can undermine epistemic agency through inference-level problems (biased, incomplete, or inadequate data and bad proxies) and enquiry-level problems (opacity, inaccessible profiles, and individualised experiences).
- The paper identifies epistemic fragmentation as a novel source of hermeneutical injustice: individuals lack access to others’ personal contexts, making it harder to share and compare experiences and to develop, uptake, and apply epistemic resources.
- Fragmentation can sustain hermeneutical injustice through three dysfunctions: blocking formation of new concepts/lacunas, preventing uptake by relevant decision-makers, and impairing application of existing concepts to one’s own experience.
- These harms disproportionately affect marginalised and disadvantaged groups; cited examples include Google Ads showing fewer high-paying jobs to women (Datta et al., 2015) and a US healthcare algorithm assigning lower health risk to Black patients (Ledford, 2019). No original quantitative results are reported.
Methods
- Conceptual/philosophical analysis; no empirical sample or data collection.
- Draws on Fricker’s epistemic injustice framework, Medina’s semantic and performative varieties, and related literature, using illustrative cases such as Carmita Wood/sexual harassment, Netflix’s *Sex Education*, Google Ads, healthcare algorithms, and social media moderation.
- Proposes a heuristic model (Fig. 1) of epistemic resource formation, uptake, application, and epistemic infrastructure, arguing that dysfunctions in infrastructure underpin hermeneutical injustice.
Limitations
- Stated: the depiction of epistemic resource creation is “highly idealised”; treatment of possible solutions lies outside the paper’s scope; the authors hope to invite more research on failures of application.
- Stated/evident: no empirical validation or measurement of epistemic fragmentation or injustice; the argument relies on conceptual analysis and illustrative/secondary examples.
- Evident: examples are drawn from other studies, fiction, and reported cases rather than original data, so prevalence and causal claims are not tested here.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures

Verification
Claims the paper makes
7 Valid · 3 Not valid · 13 Not checked · 73 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“If ML driven algorithms are considered as epistemic resources that help us to improve understanding and/or make better predictions, then the statistical nature of ML de facto prevents groups underrepresented in the data from contributing to this epistemic resource on an equal ground to dominant groups (Gebru et al., 2021)”
§ Algorithmic profiling as a source of hermeneutical injustice > Epistemological concerns with algorithmic profiling
What the sources say
- Partly supportedcites Gebru et al., 2021 sourceExperimental — accuracy not yet measured
The Introduction (p. 1) says that dataset characteristics influence model behavior and that models can reproduce or amplify biases in training data. The Composition questions (p. 6) ask dataset creators to document subpopulation distributions. Neither passage says that underrepresented groups are prevented from contributing to an epistemic resource on equal ground with dominant groups. (LLM, conf=0.82)
plainstates a 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.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“In some cases, the complex nature of the algorithm and the way it weighs large sets of different variables, make it impossible for humans, let alone laypersons, to understand why a certain algorithmic decision was made (c) (Selbst & Barocas, 2018).”
§ Algorithmic profiling as a source of hermeneutical injustice > Epistemological concerns with algorithmic profiling
What the sources say
- Supportedcites Selbst & Barocas, 2018 source
In “C. Inscrutable” (page unavailable), the paper says some decision-making rules are so complex, numerous, and interdependent that they defy practical inspection and resist comprehension. It identifies the model’s scope and sophistication, rather than a lack of expertise, as the obstacle, supporting the claim that even knowledgeable people may be unable to understand the basis of a decision. (LLM, conf=0.88)
plainstates a 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
“This effectively can be regarded as an instance of unjust hermeneutical marginalisation, in that it constitutes a situation of unequal hermeneutical participation (Fricker, 2016, p.163).”
§ Algorithmic profiling as a source of hermeneutical injustice > Epistemological concerns with algorithmic profiling
What the sources say
- Supportedcites Fricker, 2016, p.163 source
In “Preserving Patterns of Social Ignorance: Testimonial Injustice and Hermeneutical Marginalisation” (p. 5), Fricker defines hermeneutical marginalisation as a group’s not contributing on an equal footing to a shared interpretive resource. She says this puts its members at an unfairly increased risk of not being understood, supporting the citing sentence’s characterization of unequal participation as unjust. (LLM, conf=0.92)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: unclear — It is unclear whether it could 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
“Even causal explanations may be problematic, as recent work has shown (Hu & Kohler-Hausman, 2020), because they rely on flawed ontological assumptions about the social categories used for classification.”
§ Algorithmic profiling as a source of hermeneutical injustice > Algorithmic epistemic injustice > Hermeneutical injustice through failures of application
What the sources say
- Supportedcites Hu & Kohler-Hausman, 2020 source
The Abstract (p. 1) says causal fairness models assume that sex or race groups consist of people sharing an inherent trait, and calls those ontological assumptions “conceptual errors.” Section 5 (p. 9) argues that models of discrimination must also account for the social relations and meanings that constitute these groups. This supports the citing sentence’s qualified claim that causal explanations may be problematic for that reason. (LLM, conf=0.93)
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
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“In this context, it may be difficult for Adam to come forward to others about his experiences and to correctly assess them as an instance of bullying, even if he was familiar with the concept, due to a protective psychological mechanism (Gardiner, 2021).”
§ Algorithmic profiling as a source of hermeneutical injustice > Algorithmic epistemic injustice > Hermeneutical injustice through failures of application
What the sources say
- Partly supportedcites Gardiner, 2021 sourceExperimental — accuracy not yet measured
In the section labelled “8. Higher Standards” (page unavailable), Gardiner says that not realising or reporting an assault until much later can be a symptom of trauma. That supports the possibility of delayed recognition or disclosure, but the passage concerns rape, not bullying, and does not identify a protective psychological mechanism or say that negative community attitudes cause the victim’s difficulty. (LLM, conf=0.84)
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.
- Partly supported
61 more claims · 1 with their cited sources checked
Create a free account to see all 96 claims and to verify the 1 reference still open.
- It may state two separate claims as one. 17
- Probably not a substantial claim. 14
- The quote only partly states this claim. 1
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
“epistemic injustice related to AI is a growing concern.”
From A taxonomy of epistemic injustice in the context of AI and the case for generative hermeneutical erasure (2025) · cites this paper as [31]
The cited source says what the claim attributes to it.
In the citing paper
“Furthermore, [31] explain that algorithmic profiling can lead to hermeneutical injustice.”
From A taxonomy of epistemic injustice in the context of AI and the case for generative hermeneutical erasure (2025) · cites this paper as [31]
The cited source says what the claim attributes to it.
In the citing paper
“Milano and Prunkl (2025) argue that algorithmic profiling can generate hermeneutical injustice through epistemic fragmentation, making it harder for individuals to identify and conceptualise shared harms.”
From Hermeneutic Sovereignty under Generative AI: A Humanities Framework for Protecting Human Meaning-Making from Interpretive Foreclosure (2026) · cites this paper as Milano and Prunkl (2025)
The cited source says what the claim attributes to it.
In the citing paper
“Recent AI scholarship extends these concerns to algorithmic profiling and generative systems (Kay et al., 2024;Milano & Prunkl, 2025).”
From Hermeneutic Sovereignty under Generative AI: A Humanities Framework for Protecting Human Meaning-Making from Interpretive Foreclosure (2026) · cites this paper as Milano & Prunkl, 2025
The cited source says what the claim attributes to it.
In the citing paper
“The epistemic injustice literature shows why collective sense-making matters for experiences that are difficult to name within dominant categories (Fricker, 2007;Medina, 2013;Milano & Prunkl, 2025).”
From Hermeneutic Sovereignty under Generative AI: A Humanities Framework for Protecting Human Meaning-Making from Interpretive Foreclosure (2026) · cites this paper as Milano & Prunkl, 2025
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.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Algorithmic profiling as a source of hermeneutical injustice > Algorithmic epistemic injustice > Hermeneutical injustice through failures of application
“as discussed by Crear (2016).”
§ Algorithmic profiling as a source of hermeneutical injustice > References
“Crerar, C. (2016). Taboo, hermeneutical injustice, and expressively free environments.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Algorithmic profiling as a source of hermeneutical injustice > Introduction
“Symons and Alvaro (2022) link the emergence of epistemic injustice—both hermeneutical and testimonial—to algorithmic opacity.”
§ Algorithmic profiling as a source of hermeneutical injustice > References
“Symons, J., & Alvarado, R. (2022). Epistemic injustice and data science technologies.”
Citations
Cites
Black Feminist Thought: Knowledge, Consciousness, and the Politics of Empowerment.
Brewer, R. M., & Collins, P. H. (1992) · Contemporary Sociology · doi:10.2307/2074808 · Cited by 19,445
Epistemic Injustice: Power and the Ethics of Knowing
Fricker, M. (2007) · doi:10.1093/acprof:oso/9780198237907.001.0001 · Cited by 9,198 · cited 5×
“For example, it has been argued that women have a different expressive style when it comes to ethical judgments, which is sometimes marginalised as morally immature (Fricker, 2007, p.160), thereby excluding them from contributing to the hermeneutical system.”
InfluentialThis bridge called my back : writings by radical women of color
Moraga, C. L., & ANZALDÚA, G. E. (2015) · Cited by 4,964
Black looks: race and representation
(1992) · Choice Reviews Online · doi:10.5860/choice.30-2391 · Cited by 4,673 · cited 1×
“These include Moraga and Anzaldúa (1981), Hooks (1992), Alcoff (1996), and Collins (2000).”
Doyle, T. (2017) · The Information Society · doi:10.1080/01972243.2017.1354593 · Cited by 3,071 · cited 3×
“As Cathy O’Neil (2016) has described in a variety of settings, this imbalance often results in dismissal of the contextual evidence available to individuals: “An algorithm processes a slew of statistics and comes up with a probability that a certain person might be a bad hire, a risky borrower, a terrorist, or a miser…”
InfluentialBarocas, S., & Selbst, A. D. (2014) · doi:10.2139/ssrn.2477899 · Cited by 3,048 · cited 1×
“These and other types of inference-level problems are extensively researched in the fairness literature, in particular the literature on non-discrimination law and computer science (see e.g. Barocas & Selbst, 2016; Wachter, 2020).”
Gebru, T., Morgenstern, J., Vecchione, B., Vaughan, J. W., Wallach, H., Daumé, H., & Crawford, K. (2018) · arXiv · doi:10.48550/arxiv.1803.09010 · Cited by 2,474
Gebru, T., Morgenstern, J., Vecchione, B., Vaughan, J. W., Wallach, H., III, H. D., & Crawford, K. (2021) · Communications of the ACM · doi:10.1145/3458723 · Cited by 2,469 · cited 1×
“…of ML de facto prevents groups underrepresented in the data from contributing to this epistemic resource on an equal ground to dominant groups (Gebru et al., 2021), simply because they are not equally captured in the data sets used to train the predictive algorithms, or because the quality of…”
Black feminist thought : knowledge, consciousness, and the politics of empowerment
Collins, P. H. (2000) · Routledge eBooks · Cited by 2,393
(2016) · Cited by 2,334
Research works printed but not linked: 2
- Alcoff, L. M. (1996). The problem of speaking for others who can speak? Authority and critical identity. In R. Wiegman (Ed.), J Roofand. University of Illinois Press.
- Véliz, C. (2021). Privacy is power. Melville House.
Cited by
Epistemic Injustice in Generative AI
Kay, J., Kasirzadeh, A., & Mohamed, S. (2024) · Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society · doi:10.1609/aies.v7i1.31671 · Cited by 76
Two types of AI existential risk: decisive and accumulative
Kasirzadeh, A. (2025) · Philosophical Studies · doi:10.1007/s11098-025-02301-3 · Cited by 115
Mollema, W. J. T. (2025) · AI and Ethics · doi:10.1007/s43681-025-00801-w · Cited by 24
“…related to machine learning models’ epistemic opacity [20], algorithmic classification systems ’ discriminatory automation of testimonial prejudice [31], the distortion of human beliefs via the hallucinations of generative AI [5, 22], the inclusion of the global South in global AI governance [17]…”
“Does Black Box AI In Medicine Compromise Informed Consent?”
Director, S. (2025) · Philosophy & Technology · doi:10.1007/s13347-025-00860-1 · Cited by 25
Panta Rh-AI: Assessing multifaceted AI threats on human agency and identity
Plinio, S. D. (2025) · Social Sciences & Humanities Open · doi:10.1016/j.ssaho.2025.101434 · Cited by 39
Metadata
- Authors
- Milano, S., & Prunkl, C.
- Affiliations
- University of Exeter (GB)
- Utrecht University (NL)
- Corresponding author
- Milano, Silvia
- Venue
- Philosophical Studies
- Published
- 2024-02-05
- Volume
- 182
- Issue
- 1
- Pages
- 185-203
- ISSN
- 0031-8116 (ISSN-L)
- 1573-0883
- PMC
- PMC11741985
- OpenAlex
- W4391535930
- Cited by
- 38
- References
- 39
- Citations by year
- 2026: 21 · 2025: 13 · 2024: 4
- FWCI
- 17.24
- 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
- Philosophical Studies (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- Utrecht University Repository (Utrecht University) (submittedVersion, other-oa)
- Figshare (submittedVersion, cc-by)
- Open Research Exeter (University of Exeter) (submittedVersion, cc-by)
- PhilPapers (PhilPapers Foundation) (submittedVersion)
- Funders
- Wellcome Trust (223765/Z/21/Z)
- Sloan Foundation (G-2021-16779)
- Department of Health and Social Care
- Luminate Group
- Topics
- Social Sciences › Sociology and Political Science › Feminist Epistemology and Gender Studies
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Social Sciences › Arts and Humanities › Philosophy › Epistemology, Ethics, and Metaphysics
- Keywords
- algorithmic profiling, hermeneutical injustice, epistemic injustice, epistemic fragmentation, epistemic agency, epistemic resources, algorithmic harms, algorithmic mediation, social isolation, AI deployment
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 2026-10-05, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-05 (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: Algorithmic profiling as a source of hermeneutical injustice. https://app.cochat.ai/papers/10.1007/s11098-023-02095-2
- The paper
- Milano, S., & Prunkl, C. (2025). Algorithmic profiling as a source of hermeneutical injustice. Philosophical Studies. https://doi.org/10.1007/s11098-023-02095-2
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.