{"slug": "examining-variation-in-how-guided-ai-tutors-resolve-student-impasses", "title": "Examining Variation in How Guided AI Tutors Resolve Student Impasses", "summary": "A study of 20,462 student turns from 1,260 authentic sessions with a guided LLM chemistry tutor found that each additional impasse turn lowered the odds of next-turn recovery by 12.7% (AOR = 0.873, p < .001), according to the arXiv paper 2609.38346v1. Across a sample of 150 impasses, a baseline tutor gave the answer directly in 50.7% of responses, a no-direct-answer tutor asked a follow-up question every time, and the guided tutor varied its response by context. After a failed scripted question, repeating it led to recovery in 28.1% of cases versus 39.8% when the tutor addressed the student's error, and the benefit of questioning decayed with impasse depth (scripted question x depth AOR = 0.78; follow-up x depth AOR = 0.83) while addressing the error grew more beneficial (AOR = 1.14).", "body_md": "arXiv:2609.38346v1 Announce Type: new \nAbstract: When a student is stuck, a tutor faces the assistance dilemma: help given too early can hinder productive struggle, while help withheld too long leaves the student in a frustrating, persistent impasse (i.e., wheel spinning). Generative AI tutors increasingly use guardrails restricting answer-giving, yet little is known about how such tutors behave once an impasse persists. We analyze 20,462 student turns from 1,260 authentic sessions with a guided LLM chemistry tutor, identifying 6,630 impasse turns of three major types: conceptual errors, expressed uncertainty, or help-seeking. We then used these impasses to simulate three tutoring conditions to study variation in AI tutor guidance through impasses: baseline, no-direct-answer, and guided tutor. For a sample of 150 impasses, prompt specificity changed pedagogy: a baseline tutor provided the answer directly in 50.7% of responses, a no-direct-answer tutor asked a follow-up question every time, and the guided tutor responded in a wide variety of ways depending on the context. We then analyzed impasse trajectories in authentic interactions, finding that each additional impasse turn lowered the odds of next-turn recovery by 12.7% (AOR = 0.873, p < .001), and early dropouts were caught in recursive concept elicitation before reaching execution. The benefit of questioning decayed as impasses persisted (scripted question x depth AOR = 0.78; follow-up x depth AOR = 0.83), whereas addressing the student's error grew more beneficial (AOR = 1.14); after a failed scripted question, repeating it was followed by recovery in 28.1% of cases, compared with 39.8% when the tutor addressed the error instead. For learning analytics, these findings identify impasse depth and type as observable, turn-level dialogue signals that analytics can use to trigger graduated, state-sensitive assistance in real time.", "url": "https://wpnews.pro/news/examining-variation-in-how-guided-ai-tutors-resolve-student-impasses", "canonical_source": "https://arxiv.org/abs/2609.38346", "published_at": "2026-10-01 04:00:00+00:00", "updated_at": "2026-10-01 04:17:29.815318+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-research", "generative-ai", "ai-ethics"], "entities": ["arXiv", "2609.38346v1"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/examining-variation-in-how-guided-ai-tutors-resolve-student-impasses", "markdown": "https://wpnews.pro/news/examining-variation-in-how-guided-ai-tutors-resolve-student-impasses.md", "text": "https://wpnews.pro/news/examining-variation-in-how-guided-ai-tutors-resolve-student-impasses.txt", "jsonld": "https://wpnews.pro/news/examining-variation-in-how-guided-ai-tutors-resolve-student-impasses.jsonld"}}