ICLS Poster
#1079: Automated Rewriting of Lazy Prompts: How Prompt Booster Enhances ChatGPT’s Scaffolding for Academic Writing
Wed Jun 17, 4:15 PM–5:45 PM · Online
Generative AI & Large Language Models Intelligent Tutoring & Adaptive Systems Self-Regulated & Socially Shared Regulation Higher Education & Undergraduate Learning
This study introduces Prompt Booster, a Chrome extension that rewrites students’ cognitively offloading prompts into scaffolding-oriented prompts grounded in Wood et al.’s (1976) framework. Using 21 authentic prompts from the RECIPE4U dataset, we compared outputs from Original Prompts (OP), ChatGPT Study Mode (SM), and Boosted Prompts (BP). Quantitative content analysis across eight scaffolding dimensions showed that BP consistently produced more structured and process-oriented support than OP and SM (all ps < .001). Findings suggest that automated prompt rewriting can shift LLMs from content generators to instructional partners.
Authors
Shiyang Zhang, Zihao Wang, Bodong Chen