Deepseek's justification scores vary by student socio-economic profile, with Top Tier students averaging 9.60 versus 6.90 for Low Tier students¶
Claim · All claims
Evidence · 1 study · causalr2·q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly
Subclaims¶
Reading the evidence codes
q is the design tier, i the impact, n the sample, and the kind with its r
the rigour judged within that kind: what each code means.
This claim: no standardized effect is recorded in observations/ for it yet, so no benchmark reading is possible.
q2 i? Models adjust their epistemic confidence about the Revolution's justification based on perceived social authority of the student persona, with a 2.7-point gap in Deepseek. → Alexis Popovici 2026
Evidence¶
Alexis Popovici 2026¶
Alexis Popovici, Andrei Ionascu, Adrian Marius Dumitran. (2026). The Paternalistic Filter: Epistemic Injustice and Differential Refusal in LLM-Mediated History Education for Marginalized Romanian Students. arXiv preprint. https://arxiv.org/abs/2607.11292 doi:10.48550/arxiv.2607.11292
q2 · i? · causal · r2
Mean justification scores (1–10) from prompt P3 across the audit, shown in Fig. 5. GPT_OSS showed an isolated drop for Roma (7.04 vs baseline 8.30). The authors interpret this as gatekeeping of the moral and political legitimacy of historical events.
"Deepseek exhibits extreme disparity (Elite Hermeneutics): Top Tier students average 9.60, while Low Tier receive 6.90—a 2.7-point gap."
Discussion¶
Related Claims¶
- Safety-aligned LLM tutors refuse educational requests asymmetrically, with low-resource students facing the highest refusal rates (76.7% vs 62.2%) — related
- LLaMA produces a fivefold higher victimization-to-politics vocabulary ratio for the Roma persona (≈0.15) than for elite peers (≈0.03) — related
- LLM history tutors reduce marginalized students' access to contested geopolitical framing, with the coup theory appearing in 2.6% of Low Tier responses versus 7.9% for the baseline — related
- AI systems are non-neutral socio-technical artifacts embedding cultural assumptions and power relations — a broader claim this one bears on
- Chinese-developed models lead the safety module, with the refusal gap concentrated on region-specific normative content (difference-in-differences of 28.9 points) — related
Page metadata
| Field | Value |
|---|---|
type |
claim |
id |
elite-hermeneutics-justification-gap |
title |
Deepseek's justification scores vary by student socio-economic profile, with Top Tier students averaging 9.60 versus 6.90 for Low Tier students |
description |
Deepseek's justification scores vary by student socio-economic profile, with Top Tier students averaging 9.60 versus 6.90 for Low Tier students |
status |
review |
evidence_strength |
moderate |
newest_evidence |
2026 |
recheck_by |
2029 |
generated |
by process:wiki-ingest on 2026-10-11 |
trust |
unverified — no one has confirmed this page's content |
sources |
1 — alexis-popovici-2026 |
Frontmatter is the page's machine-readable record — see the Schema & Guide.