This study seeks to deepen the understanding of the direct and indirect effects of human-computer dialogic interaction programming activities, facilitated by ChatGPT, on student engagement. Data were collected from 109 Chinese high school students who engaged in programming tasks using either ChatGPT-driven dialogic interaction or traditional pair programming. A quasi-experimental analysis revealed that ChatGPT-based programming activities remarkably boost student engagement, outperforming pair programming in behavioral, cognitive, and emotional dimensions. Results demonstrated that such activities help minimize off-task behaviors, promote higher-order cognitive skills, and foster greater interest in programming. Additionally, these interactions enhance students' self-efficacy and reduce learning anxiety. The findings underscore the potential of ChatGPT-driven dialogic interaction in programming education. This study offers practical recommendations to enhance student engagement in programming learning.