Phase-Specific Perceived Difficulty and AI Scaffolding Demand in Design Thinking Final Year Projects: A Mixed-Methods Study of Diploma Electrical Engineering TVET Students at Politeknik Port Dickson

Abstract

The integration of Design Thinking (DT) into Final Year Projects (FYP) represents a promising pedagogical strategy within Technical and Vocational Education and Training (TVET) engineering programmes. However, diploma-level students may encounter phase-specific execution challenges when navigating the five DT phases of Empathise, Define, Ideate, Prototype, and Test, particularly within the ill-structured demands of FYP contexts. Concurrently, the emergence of artificial intelligence (AI) tools in educational settings raises important questions about which phases students identify as most challenging and what technology supports, they prefer. This study examined phase-specific perceived difficulty across all five DT phases and its association with preferred technology support features among Diploma Electrical Engineering students at Politeknik Port Dickson, Malaysia. A cross-sectional mixed-methods survey design was employed involving 139 students enrolled in FYP Part 1. Quantitative data were collected via Likert-scale items and multi-select instruments measuring DT satisfaction, phase-level perceived difficulty, and technology support preferences. Open-ended responses provided qualitative elaboration. Data were analysed using descriptive statistics, Cochran's Q test, chi-square tests of independence, and multivariable binary logistic regression. Findings revealed high overall satisfaction with DT implementation (M = 4.243), with Ideation (48.9%) and Prototyping (43.2%) most frequently identified as phase bottlenecks. AI Assistance (49.6%) and Mobile Application (46.0%) emerged as the most preferred support features. Logistic regression indicated that students who found Ideation and Prototyping challenging were approximately 2.5 times more likely to prefer AI Assistance, even after controlling for co-occurring phase difficulties. Mobile Application preference, by contrast, was broadly consistent across all phase-challenge profiles. These findings suggest that AI-enabled scaffolding holds particular promise for supporting students during cognitively demanding DT phases, while mobile platforms may serve as a universal delivery mechanism across all phases. Phase-level findings were further corroborated by convergent qualitative themes including idea generation difficulty, cognitive overload, and prototype execution challenges. Implications for TVET educators and DT support tool designers are discussed.