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

The DOI matches its Crossref record; 53 of the 53 references it cites were found in a registry or an index; 23 of the 55 sources its claims cite were checked for support.

Hallucination as epistemic pollution: epistemic injustice and credibility in futures-oriented science education

Open access
Küçükuncular, Ahmet · 2026 · Discover Artificial Intelligence

Abstract

In this critical conceptual review, I ask what generative artificial intelligence does to the distribution of epistemic authority in futures-oriented science education. I argue that the phenomenon usually described as hallucination is best understood not as isolated error but as epistemic pollution; that is, as a cumulative degradation of the cues by which learners judge reliability. I draw on four literature domains that have developed largely in isolation, namely technical and critical work on hallucination in natural language generation, philosophical work on epistemic injustice, futures studies scholarship on futures literacy, and science education research on future scaffolding and epistemological access. I first distinguish classes of unsupported output and specify six conditions under which such output counts as pollution rather than ordinary error, namely (1) persistence after correction, (2) circulation at scale, (3) contamination of downstream reasoning, (4) degradation of shared credibility cues, (5) opacity of provenance, and (6) the externalisation of verification costs. I then position the large language model not as an epistemic agent but as a component within a classroom credibility infrastructure, and I identify three mechanisms through which pollution may become injustice: (i) credibility displacement, (ii) hermeneutical narrowing, and (iii) unequal verification labour. Because these are conceptual hypotheses rather than demonstrated classroom effects, I state the conditions under which each would constitute an injustice rather than a harm borne equally, and what evidence would test each. I finally advance five pedagogical design principles, namely (a) provenance first inquiry, (b) collective verification, (c) epistemic role protection, (d) uncertainty discipline, and (e) plural futures safeguards. I offer these principles as regulative ideas alongside a research agenda for their empirical evaluation.

Overview

CoChat Summary

TL;DR

This critical conceptual review argues that hallucination in futures-oriented science education is best understood as epistemic pollution—a cumulative degradation of credibility cues—rather than isolated error. It specifies conditions under which such pollution may become epistemic injustice and proposes five pedagogical design principles, while stating that the mechanisms are conceptual hypotheses rather than demonstrated classroom effects.

Key findings

  • The paper distinguishes unsupported output into factual error, fabrication, unfaithfulness to source, unsupported inference, staleness, retrieval failure, and apparent contradiction; its pollution thesis applies to fabrication, unfaithfulness to source, and unsupported inference delivered in the register of warrant, plus unprovenanced factual error and staleness.
  • It specifies six conditions for epistemic pollution: persistence after correction, circulation at scale, contamination of downstream reasoning, degradation of shared credibility cues, opacity of provenance, and externalisation of verification costs.
  • It identifies three mechanisms through which pollution may become injustice: credibility displacement, hermeneutical narrowing, and unequal verification labour; four further conditions for injustice are differential model performance, differential verification position, naturalisation of dominant framings, and institutional authorisation.
  • It advances five design principles: provenance first inquiry, collective verification, epistemic role protection, uncertainty discipline, and plural futures safeguards. No quantitative hallucination figures are reported.

Methods

  • Critical conceptual review, not a systematic review or empirical study; no primary data were collected, no classroom observation, learner interview, or hallucination-rate audit was conducted, and no quantitative hallucination figures are reported.
  • Synthesises four literature domains: technical and critical work on hallucination in natural language generation, philosophical work on epistemic injustice, futures studies scholarship on futures literacy, and science education research on future scaffolding and epistemological access.
  • Searches were conducted in Scopus, Web of Science, and Google Scholar between March and July 2026; foundational philosophical and futures studies texts were not date limited, while technical and educational literature on generative systems was restricted to publications from 2020 onward, with backward and forward citation chaining from bridging works.

Limitations

  • The paper states that the relationships proposed in Sect. 5 are conceptual rather than empirically demonstrated; credibility displacement, hermeneutical narrowing, and unequal verification labour are advanced as mechanisms supported by adjacent literatures, not observed together in a futures-oriented science classroom.
  • The paper states that the classroom vignette in Sect. 5.4 is constructed, not recorded; the design principles in Sect. 6 are regulative ideas that have not been tested; and the review is critical and selective rather than exhaustive, with science education treated as a demanding test case rather than a representative one.
  • The paper also states that model capabilities and mitigation strategies change quickly, so specific technical assertions should be treated as provisional, and that the pollution metaphor may understate beneficial uses, learner resistance, and contexts where generative systems widen access. Evident from the text: the four injustice conditions and five principles remain untested, and no systematic protocol or primary data are provided.

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

Integrity signals

Registry

Crossref · 10.1007/s44163-026-02318-5

Cites

53 of 53 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

22 Valid · 0 Not valid · 23 Not checked · 123 Own results and interpretations

51 more claims · 7 with their cited sources checked

Create a free account to see all 168 claims and to verify the 1 reference still open.

See all claims — free account
  • It may state two separate claims as one. 62
  • Probably not a substantial claim. 89

Create a free account to see these claims.

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 50 references

Research works printed but not linked: 1

  1. Lewis P, Perez E, Piktus A, Petroni F, Karpukhin V, Goyal N, et al. Retrieval-augmented generation for knowledge-intensive NLP tasks. Adv Neural Inf Process Syst. 2020;33:9459–74.

Metadata

Authors
Küçükuncular, A.
Affiliations
Near East University (CY)
Corresponding author
Küçükuncular, Ahmet
Venue
Discover Artificial Intelligence
Published
2026-09-23
Volume
6
Issue
1
ISSN
2731-0809 (ISSN-L)
OpenAlex
W7214080237
Cited by
1
References
50
Citations by year
2026: 1
FWCI
5.67
Type
journal article
Language
en
Open access
Licence
Locations
  • Discover Artificial Intelligence (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Topics
  • Social Sciences › Psychology › Developmental and Educational Psychology › Educational Strategies and Epistemologies
  • Social Sciences › Education › Science Education and Pedagogy
  • Social Sciences › Education › Digital Education and Society
Keywords
hallucinations, epistemic injustice, generative AI, large language models, credibility, epistemic authority, provenance, uncertainty, futures literacy

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 and qwen3.8-omni-flash).

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: Hallucination as epistemic pollution: epistemic injustice and credibility in futures-oriented science education. https://app.cochat.ai/papers/10.1007/s44163-026-02318-5
The paper
Küçükuncular, A. (2026). Hallucination as epistemic pollution: epistemic injustice and credibility in futures-oriented science education. Discover Artificial Intelligence. https://doi.org/10.1007/s44163-026-02318-5

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.