AI-assisted TRIZ teaching with scaffolded fade-out: A quasi-experimental study in undergraduate mechanical engineering

TRIZ (Theory of Inventive Problem Solving) has earned recognition as a structured innovation methodology, yet its integration into undergraduate engineering curricula remains limited-students frequently struggle with parameter extraction and contradiction formulation, the very gates to effective TRIZ application. This study proposes a scaffolded fade-out framework that deploys an AI dialogue assistant in three progressively withdrawn phases-Proactive Guidance, Reactive Response, and On-Demand Consultation-over a four-week intervention. A quasi-experimental design (experimental group n = 42, control group n = 40) was implemented in a Mechanical Innovation Design course. Results indicate that the experimental group significantly outperformed the control group on total TRIZ modeling competence (ANCOVA F(1,79) = 18.43, p < 0.001, Cohen's d = 0.89), with large effects on parameter extraction (d = 1.15) and contradiction identification (d = 1.19). Solution innovation scores reversed (d = −0.36), which may partially reflect an anchoring effect from AI-provided inventive principles, though this interpretation requires further validation. SOLO taxonomy analysis corroborated deeper struct