Dave Diaz, Aaron Funa, Renz Alvin Gabay

ChatGPT in STEM Classrooms: Students’ Perceptions of Interest, Academic Proficiency, and Learning Independence

Introduction

Chatgpt in stem classrooms: students’ perceptions of interest, academic proficiency, and learning independence. Explore K-12 STEM students' perceptions of ChatGPT use in classrooms, examining its impact on interest, academic proficiency, and learning independence in a low-income context. Findings detail usage patterns and responsible integration.

Abstract

Purpose of the study: To provide K-12 evidence from a low- and middle-income context, this study examines how basic education STEM students use ChatGPT and how it relates to their interest, academic proficiency, and learning independence. Methodology: Design: descriptive cross-sectional survey in one large public school. Participants: 186 Grade 11–12 STEM students. Instrument: 17-item researcher-developed questionnaire, administered online during class. Tool: ChatGPT (OpenAI). Analysis: item-level frequencies and percentages; reliability and validity checks treated as supportive for a heterogeneous instrument. Main Findings: ChatGPT use was episodic and concentrated in Research and English. Students reported greater engagement, clearer understanding, and shorter assignment time. Independence gains were modest; textbook reliance declined while tutoring reliance was largely stable. Governance practices were common, including verification and paraphrase-synthesis or inspirational use. Older students emphasized efficiency and integration; younger students reported larger conceptual gains. Novelty/Originality of this study: This study contributes classroom-proximate, item-level evidence from Philippine basic education, an underrepresented K-12 setting. It characterizes selective, front-end deployment and widespread verification, offering rubric-ready handles for responsible use. It identifies grade-linked orchestration differences and connectivity-aware implications that can guide targeted scaffolds to translate efficiency into competence and independent learning.

Review

This study addresses a highly pertinent and rapidly evolving area: the integration of AI tools like ChatGPT into K-12 education. Its particular strength lies in providing much-needed empirical evidence from an underrepresented context – basic education STEM classrooms in a low- and middle-income country (the Philippines). By focusing on students' perceptions of interest, academic proficiency, and learning independence, the research offers valuable insights into the immediate impacts of ChatGPT use from the learners' perspective. The findings are not only timely but also crucial for educators and policymakers grappling with the responsible integration of generative AI into pedagogical practices. Methodologically, the descriptive cross-sectional survey of 186 Grade 11-12 STEM students in a large public school provides a solid snapshot of current usage patterns. The main findings indicate that ChatGPT use among these students is episodic and surprisingly concentrated in Research and English, rather than exclusively STEM subjects, which challenges some preconceived notions. Students reported benefits such as increased engagement, clearer understanding, and reduced assignment time, while gains in learning independence were more modest, coupled with a decline in textbook reliance. The identification of common "governance practices," including verification and paraphrase-synthesis, offers practical guidance for responsible use. Moreover, the study's contribution of classroom-proximate, item-level data from the Philippine K-12 context is genuinely novel, detailing selective, front-end deployment and widespread verification that provides "rubric-ready handles" for educators. While insightful, the study's cross-sectional and descriptive design, relying on self-reported perceptions from a single school, inherently limits the generalizability of its findings and the ability to infer causality. Future research could benefit from longitudinal designs, mixed methods approaches incorporating observed usage or performance data, and wider geographical sampling to deepen understanding of the dynamic interplay between AI tools and learning outcomes. Nevertheless, the identified grade-linked orchestration differences and "connectivity-aware implications" are particularly valuable, pointing towards the need for targeted scaffolds to effectively translate efficiency gains into genuine competence and independent learning. This work serves as an excellent foundational piece, offering critical guidance for developing responsible AI integration strategies and fostering a balanced approach to leveraging such tools in diverse educational settings.

Full Text

Comments

You need to be logged in to post a comment.