The Influence of Design and Computational Thinking on Self-Efficacy and AI Role Preferences in Mathematical Modelling

Abstract

This study investigates the relationships between Design
Thinking, Computational Thinking, and Mathematical
Modelling Self-efficacy in the context of AI-enhanced
education. Furthermore, it explores how students' levels of
DT and CT influence their preferences for different AI
roles (Tutor, Teaching Assistant, Peer, Struggling Student,
and Excellent Student) during mathematical modelling tasks.
An experimental design was employed with 26 undergraduate
participants who completed mathematical modelling tasks
assisted by AI agents. Data were analyzed using Pearson
correlations, Partial Least Squares Structural Equation
Modelling, and Mann-Whitney U tests. The results indicate
that both DT and CT are significant positive predictors of
MMSE, with DT showing a particularly strong effect.
Regarding AI role preferences, a significant difference was
found only in the Teaching Assistant (TA) role: students
with higher DT scores preferred the TA role significantly
more than those with lower scores. No significant
differences in role preferences were observed based on CT
levels. These findings suggest that while both competencies
support self-efficacy, students with high design thinking
abilities particularly value the structured yet flexible
guidance provided by a TA-style AI. This study offers
theoretical insights into the synergy of DT and CT in
mathematical modelling and provides practical implications
for designing personalized AI educational agents.