Although the use of large language models (LLMs) is considered effective in programming education, it is also known that learners may become overly reliant on the code generation capabilities of such models, potentially hindering the development of independent programming skills. In this study, we aimed to explore effective uses of an LLM not only for supporting code generation but also for enhancing learner motivation. To this end, we developed a system that provides two types of feedback: logical feedback and affective feedback, each generated through different prompts based on learners’ responses to programming tasks presented on an online learning platform. We conducted an experiment with 12 university students who were divided into two groups and asked to solve programming tasks using Ruby, a language they had not previously studied. The results suggest that participants who received affective feedback showed greater motivation to learn the language compared to those who received only logical feedback.