Understanding Factors Influencing Students Intention to Use ChatGPT for Learning Programming with Gender and IT Experience as Moderators Based on AdoptGPT Prog Conceptual Model

Despite growing interest in ChatGPT adoption in higher education, existing research lacks domain-specific models that account for the unique cognitive demands of programming education and the demographic heterogeneity among learners. This study addresses three research questions: (1) What is the relative importance of cognitive-affective factors (perceived usefulness, trust, anxiety, perceived risk) versus competency factors (AI literacy, perceived ease of use) in predicting programming students' ChatGPT adoption intention? (2) Do gender differences in technology adoption persist when examining ChatGPT specifically for programming tasks? (3) How does prior IT experience moderate the relationship between perceived benefits and adoption intention? We propose AdoptGPT-Prog, an integrated TAM-UTAUT framework incorporating eight predictors—Perceived Usefulness, Perceived Ease of Use, Trust, Anxiety, Perceived Risk, AI Literacy, Perceived Learning Value, and Hedonic Motivation—with Gender and IT Experience as moderators. Data from 486 undergraduate programming students (52.3% male, 47.7% female; mean age = 21.4 years, SD = 2.1) across five Saudi Arabian universities were analyzed using PLS-SEM. The model achieved substantial explanatory power (R²=0.714), outperforming prior ChatGPT adoption frameworks. Perceived Usefulness emerged as the strongest predictor (β = 0.312, p < 0.001), followed by Trust (β = 0.203) and Perceived Ease of Use (β = 0.187), while Perceived Risk (β=-0.156) functioned as a significant barrier. Notably, Anxiety showed no direct effect on intention (β = 0.012, p = 0.774), though gender moderated this relationship, with female students exhibiting stronger anxiety-related inhibition. Multi-group analysis revealed that females showed stronger Perceived Ease of Use effects (β = 0.279 vs. β = 0.112), while high-IT-experience students weighted Perceived Usefulness more heavily (β = 0.412 vs. β = 0.234). These findings provide empirically-grounded, actionable recommendations for differentiated instructional strategies when integrating generative AI tools into programming curricula.