Logo image
Back
Teacher-guided ChatGPT implementation in secondary school programming: evidence from a Kenyan classroom
Dissertation   Open access

Teacher-guided ChatGPT implementation in secondary school programming: evidence from a Kenyan classroom

Mutuma Ichaba
Doctor of Philosophy (Ph.D.), Drexel University
Sep 2026
DOI:
https://doi.org/10.17918/00011572
pdf
Ichaba_Mutuma_20263.32 MBDownloadView
PDF Open Access Open Access (License Unspecified)

Abstract

Educational technology ChatGPT Generative artificial intelligence Kenyan secondary education Programming education Student attitudes toward AI Teacher-guided instruction
Artificial Intelligence (AI) is beginning to enter Kenyan classrooms at a time when many secondary schools still face large class sizes, uneven access to instructional resources, and limited opportunities for individualized feedback. This study examines whether a structured, teacher-guided use of generative AI can support learning and help narrow opportunity gaps without creating new ones. The study focuses on a Grade 11 Introduction to Programming (Python) course in an urban secondary school in Nairobi. The study addresses three questions: whether participation in a teacher-guided ChatGPT intervention changed students' AI-related attitudes from pre-test to post-test; whether students' curriculum-aligned programming achievement improved over the same period, including any exploratory gender differences; and how teachers and administrators explained the observed patterns, the conditions under which they emerged, and the safeguards they considered necessary for responsible use. Using an explanatory sequential mixed-methods design, the quantitative phase relied on a one-group pre-test/post-test approach with 36 enrolled students. All 36 students had matched pre-test and post-test records within the survey dataset and within the achievement dataset; however, the two datasets did not contain a verified common identifier for student-level linkage across measures. The survey workbook used for Research Question 1 provided four prepared pre-test/post-test composites for 36 matched students: an overall mean, a positive-orientation mean, a reversed-risk mean, and a total score. Overall mean scores increased slightly from pre-test (M = 2.996, SD = 0.180) to post-test (M = 3.061, SD = 0.159), but the change was not statistically significant, t(35) = 1.711, p = .096, d_z = 0.29. The positive-orientation mean was also largely unchanged, t(35) = 0.431, p = .669, d_z = 0.07. The clearest attitudinal movement appeared in the reversed-risk composite, which increased from M = 2.802 (SD = 0.326) to M = 2.927 (SD = 0.342), t(35) = 2.036, p = .049, d_z = 0.34, suggesting a small reduction in students' perceived risk or discomfort toward AI. Broad curriculum-aligned programming and Computer Studies achievement improved from pre-test (M = 46.31, SD = 11.53) to post-test (M = 55.50, SD = 11.74), t(35) = 14.49, p < .001, d_z = 2.41. A repeated-measures ANOVA showed a significant main effect of time but no statistically reliable Time x Gender interaction, suggesting that both male and female students improved over the intervention period. The qualitative strand was used to explain why these quantitative patterns appeared. Three themes stood out. First, structured teacher modeling made AI use feel legitimate and less intimidating in the classroom. Second, verification and reflection routines pushed students to reason through AI output instead of simply accepting it. Third, the benefits were real but uneven, especially when students struggled with language precision, typing speed, time pressure, or confidence. Overall, the findings suggest that AI-enhanced instruction was most useful in this setting when ChatGPT was used as a scaffold within a teacher-guided routine rather than as a substitute for teaching, verification, or human judgment.

Metrics

2 File views/ downloads
3 Record Views

Details

Logo image