Effects of integrating a structured design thinking strategy into generative AI-supported design learning on students’ design achievement, creative self-efficacy, and problem-solving skills

The rapid integration of generative artificial intelligence (GAI) into educational settings presents a transformative opportunity for design education. However, unlocking its full potential hinges on effective pedagogical guidance. A recent study conducted at a university in China has illuminated a promising path forward, demonstrating that when GAI tools are strategically integrated within a structured design thinking (DT) framework, student learning outcomes in character design are significantly enhanced. This research offers critical insights into optimizing the use of AI in higher education, moving beyond mere tool adoption to a more purposeful and reflective learning experience.

The study, published in Frontiers in Psychology, involved 120 undergraduate students in a character design course. Participants were divided into three groups: one receiving GAI-supported learning with a structured design thinking strategy (GAI-SDTS), another using GAI without the structured strategy, and a third engaged in traditional design learning methods. The findings revealed a clear hierarchy of effectiveness, with the GAI-SDTS group demonstrating superior performance across all measured outcomes.

Key Findings Highlight Enhanced Learning Outcomes

At the forefront of the study’s revelations is the significant improvement in design achievement among students who benefited from the structured GAI-supported approach. This group not only outperformed those using GAI without a structured strategy but also surpassed students in traditional learning environments. This suggests that while GAI tools themselves can boost creative output and efficiency, their true educational value is amplified when guided by a systematic pedagogical framework.

Beyond tangible design output, the research also underscored a marked increase in students’ creative self-efficacy. Those in the GAI-SDTS group reported a stronger belief in their ability to generate novel ideas and translate them into successful design outcomes. This boost in confidence is a crucial element in fostering lifelong creative engagement and resilience in the face of design challenges.

Furthermore, the study provided compelling evidence of enhanced problem-solving skills. Students who experienced the structured GAI-supported DT approach exhibited greater confidence in their problem-solving abilities, a more proactive approach to challenges, and a stronger sense of personal control over their learning processes. This holistic improvement in problem-solving capacity is vital for navigating the complexities of modern design practice.

Contextualizing the Innovation: The Rise of GAI in Design Education

The emergence of sophisticated GAI tools like ChatGPT, Midjourney, and Stable Diffusion has rapidly altered the landscape of creative industries, including design. These tools offer unprecedented capabilities for generating visual content, exploring stylistic variations, and receiving immediate feedback based on textual prompts. For design students, this translates into the potential to visualize abstract concepts more fluidly, expedite the iterative process of trial and error, and overcome technical drawing limitations.

Historically, design education has relied on instructor-led demonstrations and extensive practice with digital tools such as Adobe Photoshop and Illustrator. While effective for building foundational skills, this traditional approach can sometimes create a disconnect between a student’s creative vision and their technical ability to execute it. The risk is that the tools themselves can become a barrier, stifling fluency and discouraging the exploration of diverse solutions. GAI, in this context, has been increasingly viewed not merely as a production instrument but as a potent learning assistant capable of supporting idea development and fostering creative thinking.

However, the study’s authors, Fenglin Song and Bing Xu, caution that the unguided use of GAI can lead to superficial engagement. Without clear instructional scaffolding, students may prioritize the aesthetic quality of AI-generated outputs over the critical thinking and problem-solving processes that underpin effective design. This can result in a passive acceptance of AI suggestions, diminishing the development of students’ independent design reasoning and their capacity to justify creative choices.

The Double Diamond: A Framework for Purposeful AI Integration

To address this gap, the researchers integrated a structured design thinking strategy based on the widely recognized Double Diamond model into the GAI-supported learning environment. Developed by the UK Design Council, the Double Diamond model provides a clear four-stage process: Discover, Define, Develop, and Deliver. This framework is instrumental in guiding students through distinct phases of divergent and convergent thinking, ensuring a thorough understanding of the problem space and fostering user-centered solutions through iteration.

In this experimental setup, the Double Diamond model served as the backbone for guiding students’ interaction with GAI tools. A specifically designed "Double Diamond-Based Design Operation Sheet" (DDDOS) was provided to students in the GAI-SDTS group. This sheet outlined stage objectives, posed self-monitoring questions, and suggested specific applications of GAI tools at each phase of the design process. This approach aimed to equip students with the ability to purposefully and reflectively utilize GAI, transforming it from a mere output generator into an integral part of a structured problem-solving journey.

Experimental Design and Methodology

The four-week intensive practical course, titled "Card Game Character Design," was conducted at a university in central China. A total of 120 second-year undergraduate animation majors participated. The course objective was to design a distinctive card game character with clear personality traits using digital tools.

The independent variable was the instructional method. The three groups were:

1. GAI-SDTS: Generative AI-supported learning with a Structured Design Thinking Strategy. Students used GAI tools (Deepseek for text generation and Dreamina for image generation) guided by the DDDOS.
2. GAI: Generative AI-supported learning without the structured strategy. Students used the same GAI tools but without the explicit DT framework.
3. TDL: Traditional Design Learning. Students relied on printed materials, online resources, and traditional digital tools like Photoshop, without GAI.

Dependent variables included:

• Design Achievement: Assessed using a rubric adapted from existing frameworks, focusing on technical skill, adherence to theme, composition, creativity, and detail quality.
• Creative Self-Efficacy: Measured using the Creative Self-Efficacy (CSE) scale, assessing confidence in generating creative ideas and performing creatively.
• Problem-Solving Skills: Evaluated using the Problem-Solving Inventory (PSI), measuring confidence, approach-avoidance tendencies, and perceived personal control.

The study employed a quasi-experimental design. For design achievement, a posttest-only control group design was used. For creative self-efficacy and problem-solving skills, a pretest-posttest control group design was implemented to account for baseline differences.

Detailed Breakdown of the GAI-SDTS Intervention

The DDDOS operationalized the Double Diamond model as follows:

• Discover Phase: Students were encouraged to explore diverse themes and visual possibilities. GAI tools were used for brainstorming character ideas through dialogue with Deepseek and for gathering inspiration and keywords for visual styles with Dreamina.
• Define Phase: This stage focused on convergent thinking, requiring students to articulate a clear character concept. Deepseek was used to refine concepts and generate prompt descriptions, while Dreamina was intentionally withheld to prevent premature visual fixation.
• Develop Phase: Students generated, compared, and refined multiple design solutions. Dreamina was utilized to create character variations and compare alternatives, with Deepseek reserved for later stages.
• Deliver Phase: The focus shifted to finalization and presentation. Dreamina was used for refining visual details and exporting the final design, while Deepseek assisted in summarizing design rationale.

Crucially, the instructor maintained a neutral facilitation role across all groups, ensuring consistency in course schedule, design requirements, and learning objectives, while avoiding intervention in students’ creative work.

Results: A Clear Hierarchy of Effectiveness

The statistical analysis confirmed the superiority of the GAI-SDTS approach across all measured outcomes.

• Design Achievement: A one-way ANOVA revealed a significant effect of instructional condition. The GAI-SDTS group (M=20.90) scored significantly higher than both the GAI group (M=18.65) and the TDL group (M=16.13). Even the GAI-only group showed improvement over the TDL group, highlighting the general benefit of GAI.
• Creative Self-Efficacy: ANCOVA, controlling for pre-test scores, showed a significant main effect. The GAI-SDTS group (adjusted M=4.25) reported the highest creative self-efficacy, followed by the GAI group (adjusted M=3.84) and then the TDL group (adjusted M=3.54). This effect was consistent across both creative ideas self-efficacy and creative performance self-efficacy dimensions.
• Problem-Solving Skills: ANCOVA also indicated a significant group effect for overall problem-solving skills. The GAI-SDTS group (adjusted M=4.44) demonstrated superior skills compared to the GAI group (adjusted M=4.04) and the TDL group (adjusted M=3.71). This advantage was evident in problem-solving confidence, approach-avoidance style, and personal control.

Discussion: The Synergy of Structure and AI

The study’s findings strongly suggest that the effectiveness of GAI in design education is not solely dependent on the technology itself but is significantly amplified by pedagogical structure. The structured DT strategy, embodied by the Double Diamond model and operationalized through the DDDOS, appears to guide students toward a more purposeful and reflective use of GAI.

The enhanced design achievement in the GAI-SDTS group can be attributed to the DDDOS’s ability to align GAI tool usage with specific design stages. This systematic approach likely helped students to leverage AI for exploration, concept refinement, and solution development without becoming overly reliant on its outputs. The study’s authors noted that while GAI can improve workflow and foster creativity, the structured DT framework ensures that these benefits are channeled into deeper design reasoning and more informed decision-making.

The notable increase in creative self-efficacy within the GAI-SDTS group can be linked to the empowerment derived from a structured learning process. By providing clear guidance and self-monitoring tools, the DDDOS likely fostered a sense of mastery and competence, enabling students to feel more confident in their ability to generate and execute creative ideas. The GAI tools, when used within this structured framework, provided opportunities for experimentation and idea validation, further bolstering this confidence.

Similarly, the improved problem-solving skills observed in the GAI-SDTS group underscore the power of DT in fostering critical thinking and adaptive strategies. The structured approach likely encouraged students to break down complex design challenges, explore multiple solutions systematically, and learn from iterative feedback, thereby enhancing their confidence and agency in tackling design problems.

Implications for Design Education and Beyond

The implications of this research for higher education design programs are profound. Firstly, it advocates for a shift from simply adopting GAI tools to strategically integrating them within established pedagogical frameworks like design thinking. This suggests that educational institutions should invest in developing curricula that embed GAI use within structured DT processes, ensuring that students are not only consumers of AI-generated content but also critical thinkers and skillful collaborators with AI.

Secondly, the study highlights the importance of externalizing the design process through tools like the DDDOS. This can empower students to self-regulate their learning, monitor their progress, and make informed decisions throughout the design journey. For educators, this offers a model for designing learning experiences that foster student autonomy and metacognitive skills.

Thirdly, the findings contribute to the broader discourse on human-AI collaboration. By demonstrating that structured guidance can mitigate the risks of passive AI use and enhance higher-order thinking, this research provides a nuanced perspective on the ethical and effective deployment of AI in educational contexts. It suggests that the true potential of AI lies not in its ability to replace human cognition but in its capacity to augment and support it when guided by thoughtful pedagogy.

Future Directions and Limitations

While this study offers valuable insights, it also points to avenues for future research. The authors acknowledge that the study primarily relied on outcome-based measures. Future investigations could benefit from qualitative and mixed-methods approaches to delve deeper into students’ learning experiences, their decision-making processes, and their interactions with GAI tools throughout the design journey. Additionally, exploring the long-term effects of this integrated approach and examining the role of instructor and peer critique in these AI-supported environments would provide a more comprehensive understanding.

The study’s findings are particularly relevant in a rapidly evolving technological landscape where generative AI is becoming increasingly accessible. As design education continues to adapt to these changes, the principles of structured pedagogical integration, as demonstrated in this research, offer a robust model for harnessing the power of AI to cultivate not just skilled designers, but also critical thinkers and lifelong learners.