Prompt specificity changes AI tutor pedagogy: a no-direct-answer guardrail alone produces uniform questioning, while a detailed guided protocol produces varied, impasse-sensitive responses¶

Claim · All claims
Evidence · 1 study · causal r2 · q2
Currency · newest study 2026 · checked against its sources 2026-10-17 (automated, 1 of 1 entry) · not reviewed by a person

Subclaims¶

q2 i? For the same 150 impasses, the baseline tutor gave direct answers in 50.7% of responses, the no-direct-answer tutor asked a follow-up question in 100% of responses, and the guided tutor used the broadest repertoire (response entropy 5.13 bits vs 3.11 and 1.25). → Ahtisham 2026

Evidence¶

Ahtisham 2026¶

Ahtisham, B., Vanacore, K., Napoli, A., Arens, J., Ionova, K., Cohn, C., Salehi, S., & Kizilcec, R. (2026). Examining Variation in How Guided AI Tutors Resolve Student Impasses. arXiv. https://arxiv.org/abs/2609.38346 doi:10.48550/arxiv.2609.38346

q2 · i? · causal · r2

Simulation study replaying a stratified sample of 150 impasse turns (n=50 per type) through three tutor conditions generated with GPT 5.6, coded with the same move set. All nine moves differed across conditions (Bonferroni-corrected p<.001; Cramér's V=.20–.56); response diversity entropies were 5.13 (Guided), 3.11 (Baseline), and 1.25 bits (No-Direct-Answer).

"The Baseline AI Tutor provided direct answers in 50.7% of responses (vs. 19.3% for the Guided AI Tutor and 0% for the No-Direct-Answer Baseline AI Tutor)."

Discussion¶

Related Claims¶

• Student impasses in LLM chemistry tutoring concentrate in the relevant-concepts planning phase, and impasse type depends strongly on problem-solving practice — related
• The association between tutor questioning and impasse recovery decays as impasses persist, while addressing the student's error becomes more beneficial — related
• Some classroom discussion approaches improve text comprehension, but gains are not uniform across approaches and more student talk alone does not produce them — related
• Next-turn recovery from impasses declines steadily as impasse depth increases, with help-seeking impasses declining most sharply — related
• Explicit no-direct-answers guardrails in teacher prompts reduced the AI's final-answer provision rate by 8.5 percentage points (16.2% to 7.7%) — related
• Model scale alone does not induce Socratic behavior: unaligned Qwen-72B achieves 0% scaffolding effectiveness and 96.7% keyword leakage — related