Analysis of Types of Inquiries in Student-AI Interaction: A case study of two CS2 tasks

Synthesis: This study analyzes the types of questions students ask generative AI systems during programming problem solving, classifying 830 interactions of CS2 students across two tasks using the Graesser et al. 18-type inquiry taxonomy via a few-shot classifier. It finds that a small subset of question types accounts for the majority of student inquiries, and that the types of questions students ask change substantially as a task progresses — evidence that student-AI interaction is not static but evolves with task demands. The work connects CS Education, Prompt Engineering, and Learning Analytics to the under-explored question of what students actually ask AI, with implications for Intelligent Tutoring design and the cognitive character of over-reliance on AI.

Key Findings

• Skewed inquiry distribution: a small subset of the 18 question types accounts for most student inquiries, suggesting students rely on a narrow repertoire of question forms when interacting with AI.
• Task-dependent evolution: the types of questions students ask change substantially as the task progresses and differ across the two programming tasks.
• Methodological contribution: a few-shot learning approach for automatically classifying student-AI interactions into the Graesser et al. taxonomy, reusable for Learning Analytics and large-scale study of Student AI Interaction.
• Context: analysis of 830 interactions from CS2 students across two programming tasks.

Why Inquiry Type Matters

Despite the importance of questioning in learning, students rarely ask enough questions in the classroom — yet they interact extensively with generative AI for learning and problem solving. Understanding what kinds of questions students pose to AI systems is therefore central to characterizing the actual learning work happening in Student AI Interaction. The authors leverage the established Graesser et al. taxonomy (18 question types) to move beyond coarse "how much AI do students use" measures toward a fine-grained account of the cognitive and epistemic character of those interactions.

The finding that students lean on a small subset of question types, and that this shifts with task progress, has direct design implications: Intelligent Tutoring and AI-learning systems can be built to invite a wider, more productive range of inquiry, and instructors can scaffold question-asking to counter a narrow, habitual interaction style. It also informs debates about Cognitive Offloading — whether students are posing deep, well-formed questions or shallow requests that delegate the cognitive work.

Connections to AI in Education

This is a measurement and characterization study in the CS Education domain. It complements research on Prompt Engineering by studying real student prompts as they naturally occur in CS2 problem solving, and it feeds Learning Analytics pipelines that aim to infer learning behavior from AI interaction logs. For instructors, the taxonomy offers a lens for noticing when students' questioning patterns are narrow or shallow and for designing interventions that broaden inquiry.

Connected Concepts

• Student AI Interaction
• CS Education
• Prompt Engineering
• LLM
• Generative AI
• Learning Analytics
• Intelligent Tutoring
• Cognitive Offloading
• Student Experience
• Higher Ed
• Metacognition

Connected Articles

• Dura LLM Cs2 — Demystify, Use, Reflect, Assess (DURA): LLM Integration in CS2
• Student LLM Interaction Taxonomy Review 2026 — Student-LLM Interaction Taxonomy Review
• Teacher Authored Prompts Student AI Dialogue — Teacher-Authored Prompts in Student-AI Dialogue
• Constructing Epistemic AI Literacy Student AI Co Programming — Constructing Epistemic AI Literacy
• Icap Cognitive Engagement LLM Agents — ICAP Cognitive Engagement with LLM Agents

Citation

Amoozadeh, M., & Alipour, A. (2026). Analysis of Types of Inquiries in Student-AI Interaction: A case study of two CS2 tasks. arXiv:2608.17919 [cs.HC, cs.AI].