<p>The acquisition of programming skills is instrumental in the development of computational thinking while also fostering abilities like collaboration and creativity. Collaborative learning has been examined as one possible approach to support student engagement and learning in programming education. However, to fully realize its potential, effective implementation of collaborative learning relies on appropriate grouping strategies to enhance learning achievement. Among these strategies, differentiated grouping based on learner characteristics is considered a key method for promoting more effective collaboration. The study proposed the Peer-Adaptive-Clustering Learning Approach (PAC-Learning Approach) in a virtual environment. It was compared to traditional methods to evaluate its effect on learning achievement, perceived computational thinking (CT) skills, and design thinking dispositions. A quasi-experimental study was conducted with 66 computer science students from a national university in Taiwan, with 35 students in the experimental group adopting the PAC-Learning Approach and 31 students in the control group following a traditional peer-adaptive learning approach. The experimental results indicate that an effective collaborative learning grouping strategy was developed, revealing four different clustering methods. The findings reveal that the PAC-Learning Approach offers a significant pedagogical advantage over traditional instructional methods, as it not only significantly improves students' post-test scores but also significantly enhances their design thinking dispositions. This study confirms the effectiveness of the PAC learning approach in enhancing both college students' learning achievement and design thinking dispositions, providing evidence for its utility in supporting programming courses.</p>