Within a single LLM session, students' help-seeking shifts toward answer- and verification-seeking while explicit self-regulation modestly increases¶

Claim · All claims
Evidence · 1 study (2 entries) · associational r? · q2
Currency · newest study 2026 · checked against its sources 2026-10-16 (automated, 2 of 2 entries) · not reviewed by a person

Subclaims¶

q2 i? Help-seeking shifted away from conceptual and procedural assistance toward answer- and verification-seeking as the interaction progressed (population slope = -.40, p < .001). → Abdelghani 2026 q2 i? SRL functions shifted modestly away from request-dominated interaction toward greater explicit planning, monitoring, or evaluation (population slope = .26, p < .001). → Abdelghani 2026 (2)

Evidence¶

Abdelghani 2026¶

Abdelghani, R., Kaiser, P., Murayama, K. (2026). Changes in Help-Seeking Strategies Predict unaided Performance during AI-based Mathematical Learning. https://arxiv.org/abs/2606.28472 doi:10.48550/arxiv.2606.28472

q2 · i? · associational · r?

Turn-level mixed-effects models using normalized within-session position, estimated on the corpus of 880 substantive turns from 112 students. The authors report a population slope of -.40, p < .001, for the shift toward closure-oriented requests.

"HS shifted away from conceptual and procedural assistance and toward answer- and verification-seeking as the interaction progressed (population slope=−.40,p < .001)."

Abdelghani 2026 (2)¶

Abdelghani, R., Kaiser, P., Murayama, K. (2026). Changes in Help-Seeking Strategies Predict unaided Performance during AI-based Mathematical Learning. https://arxiv.org/abs/2606.28472. doi:10.48550/arxiv.2606.28472

q2 · i? · associational · r?

Same mixed-effects analysis of within-session position for SRL functions. The authors report a modest positive population slope of .26, p < .001, toward explicit regulation, contrasting with the help-seeking direction.

"In contrast, SRL functions shifted modestly away from request-dominated interaction toward greater explicit planning, monitoring, or evaluation (population slope=.26,p < .001)."

Discussion¶

Learner Variables¶

• Self-Regulation — outcome: instruction changes it
• Executive Function — outcome: instruction changes it

Related Claims¶

• Evidence suggests over-reliance on generative AI may negatively affect students' self-regulated learning behaviors, potentially harming retention and transfer to novel contexts — the general claim this finding bears on, with the studies behind it
• Guest-mode usage dominates lightweight help-seeking, raising an equity concern that students without metacognitive support stay at shallow answer lookup — related
• Within-session help-seeking development predicts unaided post-test performance beyond pre-test performance and static AI-use features — related
• Grade-9 students' LLM-supported mathematical interactions are strongly request-dominated, with little explicit planning, monitoring, or evaluation — related
• Within-session SRL change shows no detectable association with unaided post-test performance — related
• Students use the integrated AI assistant differently across writing stages: clarification early, verification late — related
• Low comment rewriting does not imply a smooth process: students describe the main effort as verifying generated code rather than revising prompts — related
• Students shifted AI usage toward debugging and concept explanation over the semester, with 31 of 89 reporting more debugging use — related
• Student GenAI use patterns shifted across assessments: Learning and Formatting dominated under purpose-specific use, while Feedback became the dominant use (66.7%) under open use — related
• LLM scaffolding level is associated with the student's next dialogue act, and the association persists within broad request categories — related
• LLM scaffolding level varies systematically across student request types and across assignments, but high-assistance responses remain dominant — related
• Students' help requests to the AI tutor were predominantly instrumental rather than executive, with procedural content and low verification — related
• Students' AI workflows shifted toward guided, step-by-step use and more sophisticated prompting and critical evaluation over the semester — related
• Self-efficacy, reflective thinking, and help-seeking were associated with topic understanding at the cohort level (r = 0.427, 0.369, 0.168; d = 0.479) — related
• Educators redefined AI's educational role from answer provider to learning partner and tutor supporting conceptual understanding — related
• Direct student requests for answers significantly increased the AI's provision of direct answers (26.1% vs. 12.5% baseline; r = 0.14), though the AI maintained scaffolding in most cases — related
• Robot embodiment changes verification and debugging: learners treat robot execution as perceptual evidence for hypothesis-driven revisions, while AI-Only learners rely on generic help-seeking — related
• Scaffolded task progression shifts novice reasoning from local completion toward diagnosis and systematic debugging as task demands increase — related
• The article treats metacognition and learning as a topic requiring explicit conceptual and methodological considerations — related
• Help seeking is guided by a wider variety of considerations beyond improving immediate answer accuracy than withholding answers is
• Adaptive help seeking is a self-regulatory behavioral strategy, distinct from dependent help seeking — related
• Adolescent students' stated pre-chat intentions with an AI math tutor tilt strongly toward scaffolded support rather than answer offloading — related
• In an AI-based early warning system, the predictive importance of CAM dimensions shifted over an instructional unit, with affect/metacognition self-reports leading early and cognitive artifacts gaining accuracy toward the end — related
• Verification is nearly universal among LLM users, and most students combine multiple verification strategies rather than relying on a single one — related
• Learning-oriented intentions did not significantly predict enacted in-chat behavior; only selecting final solutions predicted executive and answer-seeking requests — related
• During AI-tutor interactions, students' turns were dominated by help requests with near-absent monitoring and evaluation — related
• Adolescent digital self-regulation is enacted through a typology of four strategy types differing in executive effort versus external scaffolding — related