Help-Seeking Behaviors in Artificial Intelligence - Mediated Programming: Impacts on Performance and Computational Literacy
Year: 2026
Type: article
Abstract: Large language models (LLMs) represent a disruptive innovation in programming education, acting as both a supportive tutor and a mechanism for cognitive offloading. This multi-method study investigated how specific student engagement patterns with a custom, course-provided LLM chatbot influenced learning performance and computational literacy among novice programmers (N = 87). Through categorical content analysis of 1,024 chatbot interaction sessions, student behaviors were classified into adaptive (learning-oriented) and maladaptive (task-completion) help-seeking strategies. Quantitative analysis revealed a mutual suppression effect, demonstrating that while aggregate usage volume was an unreliable predictor of success, the type of engagement was highly significant. Findings indicated that task-completion behaviors negatively predicted midterm exam scores and computational literacy gains, whereas learning?oriented sessions positively correlated with exam performance. Additionally, a randomized controlled trial (n treatment = 39, n control = 48) demonstrated a conditional effect: a context-aware LLM significantly improved learning performance when controlling for these usage patterns, acting as a performance multiplier. Consequently, this study advocates for a pedagogical shift from managing AI access to cultivating AI literacy, ensuring students utilize generative technology as a scaffold for cognitive development rather than a substitute for it.
Author: Fudge, Michael A.
Institution: Syracuse University
Language: English
Subfield: Computer Science Applications
Field: Computer Science
Domain: Physical Sciences
SDG: Quality education
Open Access status: green
OpenAlex ID: W7165990495
