Exploring collaborative problem-solving in AI-assisted robotics education

Abstract

Prompt engineering (PE) involves creating, improving, and deploying prompts to improve large language model (LLM) performance. Despite its increasing importance, many students lack skills in this area. Although frameworks for prompt design and process exist, theory-based instructional tools for improving prompt engineering skills are uncommon for developing collaborative problem-solving (CPS) skills in Robotics Education. This study adopted a PE strategy in AI-assisted robotics education, particularly for developing CPS. PE has four scaffolding stages: Alignment, Adaptation, Autonomy, and Innovation, and was analysed through four PE codes: Alignment Specification, Adaptive Refinement, Auditing, and Innovation. This study adopted a non-randomised, multi-site, criterion-group comparative design, with 243 undergraduate students categorised according to their ability group. This study used epistemic network analysis (ENA) to identify the co-occurrences of PE and CPS skills and examine the differences in conversation patterns among ability groups. The findings revealed distinct PE–CPS discourse configurations across ability groups. These results underscore the pedagogical value of scaffolding in developing PE skills and fostering students’ productivity in a synergetic learning environment.