Abstract

Design education attracts students with highly diverse visual spatial abilities, prior experiences, and preferred learning styles. However, most design programs continue to place all first-year students into the same foundational courses, which can lead to cognitive overload for beginners and insufficient challenge for advanced learners. This dissertation investigates how the use of artificial intelligence (AI) can support more personalized placement by integrating diagnostic sketching assessments with validated learning-style measures. The study collected 101 primary sketches from students who are interested in the field of design. The dataset is expanded to 600 images using synthetic augmentation to ensure balanced and enough representation across three sketching skill levels: Strongly Developed, Developed, and Underdeveloped. A learning-style questionnaire was also administered to classify students as Visual, Auditory, or Kinesthetic learners. Several deep-learning models were trained to classify sketching ability, and their performance was compared using accuracy, sensitivity, specificity, precision, and F1-score. The findings show that InceptionV3 achieved the highest performance across all metrics, demonstrating that AI model can accurately classify sketching ability from student drawings. InceptionV3 is particularly strong for tasks involving complex shapes and varied line quality, characteristics that are central to sketch analysis. Its ability to extract rich and diverse features from sketches explains why it emerged as the best model for classifying sketching proficiency. learning-style analysis revealed meaningful patterns: Visual learners were more common in the Strongly Developed group, Kinesthetic learners appeared across all sketching levels, and Auditory learners were more frequently found in the Underdeveloped group. These results indicate that learning style provides valuable context for understanding sketching performance and can enhance the accuracy of personalized placement decisions. The study contributes to knowledge by showing how cognitive preferences and vi- sual spatial skills can be combined to create richer learner profiles. It contributes to practice by presenting a data-driven placement model that can guide students toward appropriate prepara- tory modules, bridging courses, or advanced pathways. Limitations of the current research include the use of a single cohort, the subjectivity of sketch evaluation, and the constraints of self-reported learning-style data. Future research should expand the dataset, incorporate additional behavioral indicators, and examine the long-term impact of personalized placement on student development.

Publication Date

5-2026

Document Type

Thesis

Student Type

Graduate

Degree Name

Professional Studies (MS)

Department, Program, or Center

Graduate Programs & Research

Advisor

Ioannis Karamitsos

Comments

This thesis has been embargoed. The full-text will be available on or around 5/3/2027.

Campus

RIT Dubai

Download

Share

COinS