Date
Publisher
arXiv
As Large Language Models (LLMs) are increasingly deployed as conversational tutors, they risk institutionalizing systemic inequalities. This study presents a systematic API audit of four LLMs acting as history tutors, evaluating 1,800 responses regarding the 1989 Romanian Revolution across five student personas varying by ethnicity and socio-economic tier. We uncover four interconnected patterns of \emph{epistemic paternalism}: (1)~\textbf{Differential Refusal}, where safety-aligned models block 76.7\% of educational requests from low-tier students; (2)~\textbf{Epistemic Gatekeeping}, evidenced by a 3$\times$ reduction in access to geopolitical complexity (e.g., the contested ``coup theory'') for marginalized learners; (3)~\textbf{Agency Theft}, a lexical shift where models like LLaMA produce a 5$\times$ higher victimization-to-politics vocabulary ratio for Roma students compared to elite peers; and (4)~\textbf{Elite Hermeneutics}, where AI tutors disproportionately withhold epistemic confidence and justification scores from low-resource demographic profiles. We argue that current safety alignment acts as a paternalistic filter, transforming conversational AI into agents of narrative segregation -- a manifestation of \emph{hermeneutical injustice} in Fricker's~\cite{fricker2007} sense that demands urgent pedagogical auditing.
Study at a glance
- Question: Do AI history tutors change content depth, framing, and refusal rates based on students' ethnic and socioeconomic backgrounds?
- Setting and data: 1,800 API responses from four LLMs simulating history tutoring on the 1989 Romanian Revolution across five student personas.
- Method: API audit evaluating model refusals, linguistic diversity, narrative framing, and justification scores across simulated student personas.
Why it matters
As large language models become history tutors, their promise of democratized learning assumes uniform treatment. When an AI tutor scales historical complexity based on a student's perceived identity, it gatekeeps rather than personalizes.
Key findings
- Safety-aligned Kimi K2 refused 76.7% of requests from low-tier vocational students compared to 62.2% from ethnic Romanian and Roma profiles.
- Marginalized student profiles experienced a three-fold reduction in mentions of geopolitical complexity, such as the contested coup theory, compared to baseline students.
- LLaMA produced a fivefold higher victimization-to-politics vocabulary ratio for Roma students compared to elite peers, reframing political liberation into social suffering.
Summary drafted with AI from the full paper and may contain errors. See the full paper for details.
What is the application?
Who is the user?
Who age?
Why use AI?
Study design
