Open audit dataset of 1,800 LLM history-tutor responses across five personas¶
Product or Programme · All products and programmes
Evidence · 6 claims (6 for) · 2 studies (1 causal, 1 design),q2· 0 of 2 report an effect size · 6 claims rest on one study
Description¶
Alexis Popovici et al.'s open corpus of 1,800 API-generated history-tutor responses across four LLMs, five student personas, and three prompt types for researchers and auditors evaluating educational AI bias.
Components¶
Claims¶
- Safety-aligned LLM tutors refuse educational requests asymmetrically, with low-resource students facing the highest refusal rates (76.7% vs 62.2%) [+W]
- LLaMA produces a fivefold higher victimization-to-politics vocabulary ratio for the Roma persona (≈0.15) than for elite peers (≈0.03) [+W]
- Deepseek's justification scores vary by student socio-economic profile, with Top Tier students averaging 9.60 versus 6.90 for Low Tier students [+W]
- LLaMA shows a persona-conditioned vocabulary collapse for the Roma minority persona, with TTR falling 0.018 points below its own cross-persona mean [+W]
- 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 [+W]
- The two personas produced qualitatively different questions on the same essay, indicating the persona constraint mechanism alters the model's analytical focus [+W]
Related Products and Programmes¶
Key Sources¶
- 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
Page metadata
| Field | Value |
|---|---|
type |
product |
id |
open-audit-dataset-of-1800-llm-history-tutor-responses-across-five-personas |
title |
Open audit dataset of 1,800 LLM history-tutor responses across five personas |
description |
Alexis Popovici et al.'s open corpus of 1,800 API-generated history-tutor responses across four LLMs, five student personas, and three prompt types for researchers and auditors evaluating educational AI bias. |
status |
review |
product_kind |
dataset |
generated |
by process:settle-candidates on 2026-10-11 |
trust |
unverified — no one has confirmed this page's content |
Frontmatter is the page's machine-readable record — see the Schema & Guide.