>

>

>

High School Computer Science Students' Use of AI Coding Tools and Debugging Confidence: A Survey of 31 Students

High School Computer Science Students' Use of AI Coding Tools and Debugging Confidence: A Survey of 31 Students

Ido Goren

https://doi.org/10.67698/pjpcr.2026.0215

Abstract

Generative artificial intelligence (AI) coding tools are now part of how many high school students learn to program, but little is known about how students themselves describe their use of these tools and how it relates to their confidence in debugging. This study surveyed 31 high school students (grades 9–12) who were taking or had taken a computer science course. The survey combined multiple-choice items on tool use, five-point Likert items on debugging confidence and attitudes toward AI, and open-ended questions. About 61% of students used AI coding tools at least weekly, and 52% used them several times a week or every time they coded. The most common purposes were getting explanations (61%), understanding error messages (61%), debugging (52%), and learning new concepts (52%); 35.5% also used AI to generate code. Debugging confidence, measured as the mean of five Likert items, was not significantly related to frequency of AI use (Spearman ρ = .06, p = .73), and a chi-square test on a single error-message item also found no association, χ²(1, N = 31) = 0.82, p = .37. With 31 students, only fairly large relationships (|r| of about .48 or more) could be detected reliably, so small associations cannot be ruled out. Most students (81%) agreed that AI tools helped them understand their mistakes, 71% preferred to try solving a bug themselves before using AI, and 77% agreed they could write original code without AI, yet only 23% agreed they could fix bugs in unfamiliar code. Open-ended answers from 18 students described AI as useful for explanations and error finding, and on a separate concerns question, 8 of 18 substantive answers raised overreliance, overuse, or loss of problem-solving skill. These exploratory findings suggest that students use AI mainly as an explainer and error-finder, and they point to the need for larger studies that follow debugging skill over time.

100%
Bind a PDF file to preview.

Princeton, New Jersey, United States
Published and Managed by The Princeton Journal of Precollegiate Scholarship Inc.
ISSN: 3143-8423
DOI: 10.67698

Copyright © Princeton Journal of Pre-Collegiate Research. All rights reserved

PJPCR is independently operated and is not affiliated with Princeton University or any of its colleges, departments or programs.

Princeton, New Jersey, United States
Published and Managed by The Princeton Journal of Precollegiate Scholarship Inc.
ISSN: 3143-8423
DOI: 10.67698

Copyright © Princeton Journal of Pre-Collegiate Research. All rights reserved

PJPCR is independently operated and is not affiliated with Princeton University or any of its colleges, departments or programs.

Princeton, New Jersey, United States
Published and Managed by The Princeton Journal of Precollegiate Scholarship Inc.
ISSN: 3143-8423
DOI: 10.67698

Copyright © Princeton Journal of Pre-Collegiate Research. All rights reserved

PJPCR is independently operated and is not affiliated with Princeton University or any of its colleges, departments or programs.