Research by Qian Huang & King Wang Poon

Literature Review: Foundations of AI Literacy

The integration of artificial intelligence into education is a rapidly evolving field characterized by both significant opportunities and profound challenges. Research in this area spans two primary domains: "learning with AI," which involves using AI to augment teaching and learning processes, and "learning about AI," which focuses on educating students about the principles, applications, and societal implications of AI itself. This study falls squarely within the latter, exploring how students learn about AI by actively creating it. The literature on AI in education highlights a growing consensus that simply teaching students to use AI tools is insufficient for preparing them for an AI-shaped world. There is a pressing need to develop frameworks that promote deeper forms of engagement and critical thinking.

A key concept emerging from recent scholarship is AI literacy, which has been conceptualized as a multi-faceted competency encompassing awareness of AI systems, the ability to use and adapt them for specific tasks, critical evaluation of their outputs, and ethical understanding of their implications. This paper aligns directly with this vision, as the process of building custom GPTs necessitates that students engage with all four components of AI literacy simultaneously.

Theoretical Framework: Tool-to-Teammate Strategy

The pedagogical framework guiding the course is built upon a robust, multi-theoretical foundation that explains the cognitive, motivational, and social dimensions of the "Tool-to-Teammate" strategy. This framework integrates four complementary theories: Self-Determination Theory (SDT), Constructionism, the Technological Pedagogical Content Knowledge (TPACK) framework, and Distributed Cognition.

SDT explains the motivational core, fostering intrinsic motivation through autonomy, competence, and relatedness as students create custom GPTs. Constructionism explains the cognitive power of "making," as students externalize understanding by building functional AI artifacts. TPACK provides a map of knowledge domains needed for integrating technological, pedagogical, and content knowledge in AI-rich contexts. Distributed Cognition conceptualizes the custom GPT as an integral part of the student's extended cognitive system, functioning as a cognitive partner rather than just a tool.

Methodology: Multi-Case Study Approach

To investigate the transformation of students from passive AI users to active creators, this research employed a qualitative multi-case study methodology. This approach is particularly suited for exploring complex phenomena within their real-life contexts and is ideal for generating rich, detailed insights into the processes and outcomes of innovative pedagogical strategies. The unit of analysis was the 'Creating the Frontier of No-code Smart Cities' course at the Singapore University of Technology and Design (SUTD), with three distinct student-built GPT-based projects serving as the focal cases for in-depth examination: the Interview Companion GPT, the Urban Observer GPT, and Buddy Buddy.

Data collection was triangulated from multiple sources including project reports, design specifications, final deployed GPT tools, interviews with student project leads and instructors, and reflective essays. The analytical process involved both within-case and cross-case analysis, focusing on the interplay between Design, AI, and Domain Knowledge, contrasting "Before" (generic tool) with "After" (customized teammate) states.

Cross-Case Synthesis: The Trilingual Triad Synergy

A cross-case synthesis of the three projects reveals a powerful and consistent learning model driven by the dynamic interplay between Design, Artificial Intelligence, and Domain Knowledge, conceptualized as a "Trilingual triad." This model demonstrates how these three elements are deeply interconnected components of a unified system of learning, orchestrating synergy rather than isolated mastery.

Domain Knowledge informs AI Architecture: Successful AI creation requires deep understanding of domain rules to codify them into the AI's logic (e.g., interview rules, urban planning theories). This forces metacognition, transforming tacit knowledge into explicit criteria.

Design Bridges Theory and Practice: Thoughtful design mediates AI's capabilities for educational utility. Interfaces, workflows, and interaction patterns channel AI's power toward specific pedagogical goals (e.g., multimodal interface, "Friendly Tutor" persona, interactive onboarding). Good design ensures AI augments, rather than replaces, human learning.

AI Empowers Agency: The synergistic process ultimately empowers student agency. By shaping AI to fit their needs, students gain competence and autonomy, becoming conscious and critical architects of technology. This aligns with Self-Determination Theory and fosters a profound sense of ownership and intrinsic motivation.

Conclusion: Shaping the Future of AI Education

The 'No-code Smart Cities' course at SUTD offers a vital glimpse into a future of education where students are empowered as creators and critical partners in the age of AI. By embracing a pedagogy grounded in the synergy of domain knowledge, thoughtful design, and AI capabilities, and guided by a robust multi-theoretical foundation, educators can foster a generation of learners who are not just prepared to use AI, but to shape it responsibly and effectively.

The "Trilingual triad" model offers a replicable framework for integrating no-code AI development into higher education, cultivating a sophisticated form of AI literacy that extends beyond operational proficiency to encompass critical evaluation, ethical consideration, and creative application. This approach fosters a partnership between human ingenuity and artificial intelligence, positioning students as co-designers and collaborators.