Students' AI workflows shifted toward guided, step-by-step use and more sophisticated prompting and critical evaluation over the semester¶

Claim · All claims
Evidence · 1 study · associational r1 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? AI-guided step-by-step workflows nearly tripled (7 to 20), concept explanation requests more than doubled (12 to 28), and markers of sophistication — prompt constraint setting (8 to 15), explicit skepticism (4 to 11), and diagram-interpretation awareness (1 to 8) — all grew. → Shuang Geng 2026

Evidence¶

Shuang Geng 2026¶

Shuang Geng, Helen Lallos-Harrell, Jiya Ashar, Thomas J. McKenna, Annwesa Dasgupta, Caleb Farny, and Emma Lejeune. (2026). Structured AI Demonstrations and Student LLM Use in Engineering Mechanics: Study Design and Preliminary Results. https://arxiv.org/abs/2607.28710 doi:10.48550/arxiv.2607.28710

q2 · i? · associational · r1

Machine-coded free-text workflow responses (pre n=100, post n=99) from the adjudicated multi-model LLM coding pipeline; each response received exactly one overall process-structure code. "AI-guided step-by-step workflows nearly tripled (7 to 20)" while non-use fell 16 to 9.

"AI-guided step-by-step workflows nearly tripled (7 to 20) as non-use fell (16 to 9)."

Discussion¶

Related Claims¶

• Students shifted AI usage toward debugging and concept explanation over the semester, with 31 of 89 reporting more debugging use — related
• Within a single LLM session, students' help-seeking shifts toward answer- and verification-seeking while explicit self-regulation modestly increases — related
• Participants used the Flowcode LLM across all stages of their design process, most frequently asking for code explanations — related
• More students embraced LLMs for debugging from Spring to Fall 2025 despite declining course enrollment — related
• Student dialogue prompted educators to reconstruct teaching beliefs about assessment, debugging instruction, and active learning — related
• Student appraisals became more critical over the semester, with actionable-support themes falling from 67% to 43% — related
• Verification behavior lagged rising AI use: minimal or conditional verification more than doubled (5 to 12) over the semester — related
• Students describe concrete verification behaviors for GenAI explanations, treating evaluation as an iterative filter rather than a single accept-or-reject judgment — related
• By the final survey wave, using LLMs for debugging and understanding concepts was nearly universal among surveyed students — related
• Curiosity about AI shifted between pre- and post-survey, from 71% to 50%, which the authors interpret as a move from diffuse curiosity to concrete critical engagement — related