Exploring trust in artificial intelligence in a problem-based computer programming course

Generative artificial intelligence is increasingly used in programming education, but evidence concerning its relationship with achievement and learner trust remains limited in Nigerian higher education. This mixed-methods, quasi-experimental study examined whether integrating Google Gemini into problem-based learning (PBL) was associated with undergraduate students’ programming achievement and trust in AI. Sixty-two students from intact classes participated in a pre-test–post-test design: 29 received Gemini-enhanced PBL, and 33 received PBL without AI. Programming achievement and multidimensional trust were assessed before and after the intervention, while eight experimental-group students provided explanatory interview accounts. Quantitative data were analysed using ANOVA and ANCOVA, supplemented by HC3 robust estimation and bias-corrected bootstrapping; interview responses were coded thematically. The experimental group recorded higher post-test achievement than the control group. The robust adjusted mean difference was 16.42 points, 95% CI [14.16, 18.67]. Post-test trust scores were also higher in the Gemini-enhanced condition, while gender and the instructional condition × gender interaction were not statistically significant. Participants described Gemini as useful for problem clarification, algorithm development, debugging, collaboration, self-directed learning and immediate feedback; trust was associated with transparent prompts, responsible data use and human accountability. The findings suggest that structured Gemini support may complement PBL.