Bridging Technology and Creativity: A Study on AI Adoption in Product Design Education

Types of AI Applications in Design Education

AI-Generated Content tools such as ChatGPT, Midjourney, Stable Diffusion, DALL·E, and Kinectt have become central in supporting design education [6]. These tools generate text, images, code, and 3D models, which enable students to experiment with concepts more rapidly than traditional methods [7]. Generative AI, in particular, enhances ideation and prototyping by expanding the range of creative options available to students. Similarly, AI-powered platforms streamline content generation, translation, and localization, making course preparation more efficient [8]. However, while these technologies enable rapid iteration and efficiency, concerns remain about potential over-reliance, which could limit the development of students’ independent problem-solving skills [9].

Pedagogical Frameworks and Learning Strategies

The integration of AI in design education also requires appropriate pedagogical frameworks. Studies emphasize that AI aligns with Education 4.0, supporting adaptive learning, flexible course delivery, and skills development for future industry needs [3, 4]. Frameworks such as first-principle-based coursework have been proposed to balance AI support with foundational reasoning, thereby ensuring that critical thinking remains central to student learning [9]. Moreover, AI’s predictive capabilities can help align learning outcomes with industry demands by forecasting emerging design trends [8]. These studies highlight that AI should not be viewed as a substitute but as a complement to traditional methodologies, enhancing efficiency while preserving core educational objectives [10].

Ethical and Critical Implications of AI Adoption

A significant body of literature also addresses ethical concerns in AI integration. Bias embedded in AI algorithms poses a risk of perpetuating inequities in design education if not carefully mitigated [11]. Students must therefore be equipped with tools to identify and challenge biases, ensuring inclusivity in their outputs. Additionally, overreliance on AI raises questions about originality, with risks of plagiarism and reduced student creativity [6]. The need for continuous upskilling of both students and instructors is another recurring theme, as rapid technological advances require sustained adaptability [9]. Together, these concerns underscore the importance of embedding AI literacy and ethics into the curriculum alongside technical competencies [10, 11].

Evaluation Rubric for Functional AI Tools

To provide methodological rigor in assessing AI applications, this study introduces a two-dimensional rubric that evaluates AI tools along the axes of Interactivity () and Generativity (), as depicted in Figure 2.

Symbolic Representation:

Each AI tool is represented as a coordinate pair:

Composite score (CS) is computed as [14]:

where represents the level of interactivity and represents the level of generativity. Equal weights () are used to balance both dimensions. The rubric matrix of this assessment method is summarized in Table 1.

Concept Generation

In the ideation phase, AI tools such as ChatGPT and Midjourney assist students in generating creative concepts. These tools facilitate brainstorming by providing diverse ideas and design alternatives based on text prompts, enabling students to explore various possibilities before advancing to the design phase. The use of generative AI in concept development has been shown to expedite the ideation process and enhance creativity [15]. Figure 3 illustrates the conceptual framework for using AI tools in the ideation phase of product design. The inverse pyramid diagram highlights the stages from Generated Content (GC), AI-supported GC, to Outcome-Based GC. It demonstrates how AI tools such as ChatGPT and Midjourney assist students in generating creative concepts by providing diverse ideas and alternatives based on user prompts. This approach allows for rapid exploration of various design possibilities before advancing to the implementation phase.

Design Implementation

During the design phase, AI tools are integrated to aid in creating and refining design models. For instance, generative AI systems like Stable Diffusion and Midjourney generate 2D or 3D designs based on user inputs. Additionally, tools like ChatGPT serve as co-pilots in generating code for specific tasks, enhancing efficiency and creativity. The design process includes

• Ideation and sketching: AI tools generate initial sketches and concepts based on textual descriptions provided by students.

• Prototyping and simulation: AI tools facilitate rapid prototyping by generating CAD (Computer Aided Design) models or assisting in coding for hardware components.

• Testing and refinement: Students test their designs in virtual environments, allowing for rapid iteration and refinement.

The integration of AI in design implementation has been demonstrated to accelerate the design cycle and improve design quality [16]. Figure 4 below presents the structured workflow for implementing AI tools in the design process, including Concept Generation, Design, Implementation, and Operation [12]. It outlines how AI tools support each stage by facilitating ideation, generating models, assisting in prototyping, and enabling virtual testing. The diagram emphasizes the iterative nature of the design process, where students continuously refine their designs based on feedback and simulation results provided by AI tools.

Case Examples of AI Tool Integration in Student Projects

To complement the survey data, anonymized student projects were analyzed to illustrate how AI tools directly influenced the design process. Three representative cases are presented below.

Case 1: Concept Generation with ChatGPT

Input (student prompt): “Generate three possible design concepts for a sustainable desk lamp suitable for small student apartments, focusing on affordability and ease of assembly.”

AI output (ChatGPT):

Foldable desk lamp with interchangeable LED panels powered by USB.

Modular wooden lamp kit using snap-fit joints for easy assembly without tools.

Desk lamp with integrated pen holder, fabricated from recycled PLA filament.

Student outcome: Students selected the modular wooden kit concept, developed sketches, and later prototyped using laser-cut plywood. The AI suggestions were credited with accelerating brainstorming and broadening creative options.

Case 2: Visual Rendering with Midjourney/Stable Diffusion

Input (student text prompt in Midjourney): “A modern ergonomic chair designed for small co-living spaces, minimalistic style, rendered in natural wood and soft fabric.”

AI output (image renderings): AI generated four variations of the chair, showcasing different textures, lighting, and proportions.

Student outcome: The rendered images were used as initial references in CAD modeling. Students highlighted that AI-generated visuals helped them refine proportions quickly and communicate ideas effectively during peer critiques.

Case 3: Concept generation for product design for Cycling Device that can be enjoyed by parents and kids. In this example, the complete experiment is illustrated from concept generation, design iteration, implementation, operation, and reflection, which cover six main steps.

Step 1: Problem Definition. Prompts were provided to the AI to fine-tune the problem definition. The following statements were generated.

Parents often want to cycle safely and enjoyably with their children, but current solutions (child seats, trailers, or tag-along bikes) have limitations:

• Safety concerns on busy roads.

• Lack of engagement for kids during the ride.

• Limited adaptability as the child grows.

• No shared activity beyond transportation.

The goal is to design a cycling companion device that enhances interaction, safety, and fun for both parents and kids.

Step 2: Concept Generation Directions. Further prompts to AI were provided to generate possible options. Two options are listed here for illustration.

Option A. Smart Interactive Cycling Companion

A detachable module with voice intercom, lights, and sounds, allowing communication between parent and child bikes.

Features:

• Bluetooth intercom for hands-free talk.

• LED mood lighting synced to speed or terrain.

• Interactive “mission mode” for kids (e.g., collecting virtual badges by distance).

Option B. Modular “Follow-Me” E-Bike Trailer

• Converts between child trailer, cargo trailer, and solo kid bike.

• Uses a magnetic quick-link hitch for easy attachment/detachment.

• Electric assist synchronizes with the parent’s pedaling to ensure consistent speed and less strain uphill.

Step 3: Interactive design by including design preference. Figure 5 shows an initial concept sketch generated by AI. How about using a bicycle trailer? A new concept sketch generated by the AI is shown in Figure 6, where designer’s preferences were provided to AI.

Step 4: Minimum electronics design. Kids have been exposed to a plethora of electronic devices at home, including computers, smartphones, video game consoles, and various other gadgets. Consequently, the design should prioritize a minimalistic approach to electronics.

Step 5: Generative AI was used for engineering design. Autodesk Fusion 360’s AI-based generative design feature was used to model a weight-optimized chassis [17]. Traditional engineering techniques could be used to create an over-engineered chassis, but they would lead to wastage of materials and result in a bulky chassis that would be difficult to handle in a trailer form that would be towed.

Step 6: Practical Final Design. The final working prototype for the parent and child cycling together device is depicted in Figure 7.

Evaluation

The final stage involves assessing the effectiveness of AI tools in enhancing student learning outcomes. Surveys were administered to students to gather feedback on their experiences using AI tools throughout the design process. The surveys aimed to evaluate the impact of AI tools on creativity, efficiency, and overall satisfaction.

The survey was conducted among 41 undergraduate students enrolled in an Engineering Design program (ages 18–24 years). No gender identity data were collected to maintain privacy and confidentiality. All participants shared a common disciplinary background in Engineering Design.

Prior to data collection, ethical clearance was obtained from the Institutional Review Board (IRB) of the Singapore University of Technology and Design. Informed consent was secured from all participants, and anonymity was preserved. This aligns with the declaration provided in the Ethical Statement of this manuscript.

To evaluate student perceptions of AI adoption, survey responses were collected using a five-point Likert scale (Strongly Agree = 5 to Strongly Disagree = 1). The statistical analysis results are summarized in the tables (Tables 2–4).

The t-values for Questions 1–7 in Table 2 exhibit exceptionally large magnitudes, ranging from –7.00 to +28.13, signifying that the mean responses differ markedly from the neutral midpoint of the Likert scale. Collectively, these results reflect a near-unanimous consensus among participants and require interpretive depth. For Question 1 (“Shall we completely ban AI usage in the Engineering Design course?”), the negative t (–7.00) indicates strong disagreement with banning AI, confirming that students overwhelmingly view AI as an enabling rather than threatening presence in their learning. Questions 2 and 3 show very high positive t-values (20.96 and 17.75, respectively), revealing that students firmly believe AI reduces design iterations and improves design quality. These results suggest that learners regard AI as a cognitive accelerator that enhances workflow efficiency and outcome quality, though educators must remain cautious that accelerated processes do not compromise analytical reflection or verification.

For Questions 4 and 5, the t-values reach the highest magnitudes (26.72 and 28.13), demonstrating an almost unanimous intention to continue using AI tools and to explore new ones as they emerge. Such findings imply a behavioral commitment to AI-enhanced design and reflect a student culture of technological openness and curiosity. Questions 6 and 7, with t-values of 18.57 and 14.34, respectively, reveal significant satisfaction with both the process and outcomes of AI-assisted design compared with traditional design-thinking-innovation (DTI) methods. The slightly lower t for Question 7 suggests that, while students find AI-enabled processes engaging, they still differentiate between process satisfaction and the tangible quality of final products—an observation that may point to gaps in prototype validation or physical testing.

Overall, the consistently large t-values arise from high mean scores (≈4.5–4.9) and low variance (SD ≈ 0.5), indicating strong homogeneity rather than random variability. This statistical pattern confirms that students perceive AI as an integral and beneficial collaborator within the design process. However, the near-ceiling clustering of responses also limits discrimination across items, implying that future instruments should employ broader or more nuanced scales to capture subtle attitudinal variations. These outcomes should therefore be interpreted as evidence of strong endorsement of AI integration, rather than as indicators of measurable performance gain.

Beyond the inferential statistics, the descriptive distributions in Tables 3 and 4 offer deeper insights into how students deploy AI tools across the design workflow. Table 3 shows that ChatGPT or equivalent language-based systems were used by 66% of respondents, far exceeding the 34% who employed Midjourney or other image-generation tools. This discrepancy underscores the broader functional utility of interactive text-based AI for ideation, documentation, and reflection—activities that dominate engineering design education. The distribution of AI utilization across design stages further reveals that ideation accounts for 48% of total AI use, followed by sketching (23%), whereas prototyping, testing, and refinement collectively represent only ≈30%. This imbalance demonstrates that students rely on AI primarily during the early divergent phases of design, where creativity and exploration are essential, while later convergent phases that require empirical validation and material interaction remain largely manual.

Table 4 extends this observation by distinguishing between interactive and generative AI modalities. The data indicate that interactive AI dominates ideation (45%) and refinement (41%), consistent with the dialogic nature of ChatGPT and Copilot, which facilitate real-time reasoning, critique, and feedback. Conversely, generative AI prevails in sketching (66%), where visual synthesis and multimodal rendering are central, enabled by platforms such as Midjourney or Stable Diffusion. In contrast, prototyping and testing remain largely manual (≈77%), confirming that current AI technologies have limited capability in the physical and experimental domains of product development. This pattern corresponds closely with the CDIO–RI framework, where interactivity aligns with the Concept and Reflection stages, while generativity supports the Design and Implementation phases.

Analytically, the descriptive results reveal a bimodal pattern of AI engagement: interactive systems dominate tasks requiring contextual dialog and analytical reasoning, whereas generative systems dominate creative visualization. This adaptive usage suggests that students are not indiscriminately adopting AI but are developing functional discernment—deploying each tool according to its cognitive affordance. Nevertheless, the steep decline of AI usage in prototyping and testing highlights a persistent gap between digital ideation and tangible realization. Bridging this divide will require pedagogical innovations that integrate AI-assisted simulation, fabrication, and performance evaluation into later stages of the curriculum. The combined descriptive and inferential findings provide a coherent picture of strong AI acceptance, coupled with selective, context-driven application, and offer clear guidance for the next evolution of AI-enhanced product design education.

The survey results indicated that:

• 70% of students disagreed with banning AI tools from the curriculum.

• 27.5% remained neutral.

• Only 2.5% supported the ban on AI tools.

• 66% of students reported extensive usage of ChatGPT or equivalent tools.

• 34% of students utilized Midjourney or similar tools.

These findings demonstrate a positive reception toward incorporating AI tools in product design education. Further analysis of the survey data is presented in the Discussion section. The use of surveys to evaluate AI tools in education is a common practice to assess their impact on student learning and engagement [18, 19]. Table 5 illustrates the interaction between students and instructors in the classroom setting with AI tools. It shows how AI tools are utilized by students for generating project proposals, creating code, and developing designs, while instructors use AI tools to evaluate and provide feedback. The diagram highlights the bidirectional learning process, where both students and instructors continuously adapt to emerging AI tools to enhance the educational experience.

AI for Design Process and Effectiveness

The findings derived from the survey and classroom observations provide multifaceted insights into how students perceive, adopt, and apply AI tools throughout the design process. The expanded discussion below interprets the results of Tables 2–4, and relates them to the CDIO–RI framework and the conceptual models illustrated in Figures 2–4.

Benefits of AI Tools in Product Design Education

Integrating AI tools into product design education offers several notable benefits:

• Enhanced Creativity: AI tools, particularly generative models like ChatGPT and Midjourney, facilitate brainstorming and ideation by providing a wide range of concepts based on user inputs [20]. This expands students’ creative potential and enables rapid exploration of novel ideas [21].

• Improved Efficiency: Automating processes such as coding, content generation, and rendering allows students to focus more on refining their designs rather than performing repetitive tasks [22]. This efficiency leads to faster prototyping and iteration cycles, essential for effective learning.

• Personalized Learning: AI tools offer personalized feedback and suggestions tailored to individual students’ needs, helping them overcome challenges more effectively and promoting a more inclusive learning environment [23].

Challenges and Limitations

Despite the benefits, the use of AI tools in product design education presents several challenges:

• Over-reliance on AI Tools: Excessive dependence on AI tools may diminish students’ critical thinking and problem-solving skills [24]. When students rely too heavily on AI-generated suggestions, their ability to independently create and refine designs may be compromised.

• Ethical Considerations: AI systems can perpetuate existing biases present in training data. Without careful monitoring, biased outputs can negatively impact the inclusivity and quality of design solutions [25].

• Technological Constraints: The rapid evolution of AI tools requires continuous adaptation from both instructors and students. Staying updated with the latest technologies can be challenging, especially when resources are limited.

• Skill Gaps: Effective use of AI tools demands specific skills that students and educators may not yet possess. Ensuring adequate training and support is essential to harness the full potential of AI in design education.

Discussion

The results underscore the transformative potential of AI tools in product design education. While the benefits are evident, addressing associated challenges through structured pedagogical approaches is crucial. Integrating AI literacy and ethics into the curriculum can help mitigate potential drawbacks while maximizing the positive impact of AI tools on learning outcomes.

Fostering a balanced approach that combines AI-generated content with traditional design methodologies may help maintain students’ critical thinking and creative skills. Educators must continuously adapt to emerging technologies and refine their teaching strategies to meet the evolving demands of the industry.

The findings also demonstrate that AI tools such as ChatGPT, Midjourney, and Stable Diffusion are not only accepted by students but also actively integrated into their creative and technical design processes. This section provides a more comprehensive discussion by situating the results within existing research and elaborating on their pedagogical, technological, and ethical implications.