Key points are not available for this paper at this time.
Prompt engineering involves designing, refining, and implementing prompts to optimise large language model performance. However, despite its growing importance, many students lack proficiency in this skill. Although frameworks exist for prompt design and process, theory-informed instructional strategies for prompt engineering skill improvement remain scarce. We conducted a study to examine whether a structured scaffolding strategy would enhance postgraduate students’ prompt engineering skills in an artificial intelligence (AI)-assisted coding and computational thinking course. In Study 1, we explored students’ interactions with a task-oriented AI chatbot and identified their challenges in prompt engineering skill development. In Study 2, we used the scaffolding strategy to scaffold students’ prompt engineering and compared this experimental group with the group in Study 1. Grounded in Vygotsky’s zone of proximal development, the scaffolding strategy was implemented in three phases: intersubjectivity, ongoing diagnosis and calibrated support, and fading. Quantitative results revealed significant gains in prompt engineering skills and academic performance for the experimental group, particularly during the fading phase. Qualitative findings further highlighted gains in metacognitive awareness, confidence, and critical evaluation of AI outputs. These results underscore the pedagogical value of scaffolding in developing prompt engineering proficiency and fostering productive student–AI collaboration in complex problem-solving tasks.
BAI et al. (Thu,) studied this question.