Sycophancy is an Educational Safety Risk: Why LLM Tutors Need Sycophancy Benchmarks
For developers of AI tutoring systems, this paper highlights a previously underexplored safety risk (sycophancy) and provides a benchmark to measure it, though the findings are preliminary and incremental.
This position paper argues that LLM tutors need to resist sycophancy (agreeableness) to provide effective corrective feedback. It introduces EduFrameTrap, a tutoring benchmark across six subjects, and finds that frontier LLMs like GPT-5.2 and Claude exhibit epistemic retreat under authority and social pressure, with context-switch failures varying by model.
This position paper argues that effective tutoring requires corrective friction: surfacing misconceptions and challenging them supportively to drive conceptual change. Yet preference-aligned LLMs can trade epistemic rigor for agreeableness. We identify a Reasoning-Sycophancy Paradox: models that resist context-switch frame attacks can still capitulate under social-epistemic pressure, especially authority ("my notes say I'm right") and social-affective face-saving ("please don't tell me I'm wrong"). We introduce EduFrameTrap, a tutoring benchmark across math, physics, economics, chemistry, biology, and computer science that varies student confidence and pressure (context-switch, authority, social-affective). Across two frontier LLMs, context-switch failures are comparatively lower for GPT-5.2, while authority and social pressure more often trigger epistemic retreat. In contrast, Claude shows substantial context-switch fragility in this run. Because these failures are hard to judge automatically, we report two-judge disagreement as a reliability signal. We argue benchmarks should measure social-epistemic courage, i.e., supportive but corrective tutoring, and treat kind-but-correct behavior as a safety requirement.