Lingyi Kong
Adjunct Professor
Parsons School of Design
The underlying logic of generative AI—structured prompts, output predictability, and system feedback—is not unlike the foundations of critical design education: both rely on linguistic construction, syntactic control, and traceable iterations. This project explores that shared grammar as a pedagogical entry point, training students to treat generative models not just as tools, but as systems of meaning-making.
The framework introduces a “backward design” pedagogy in which students begin with AI outputs and work backward to decode the system’s structural assumptions. They analyze how prompt phrasing affects visual/linguistic output, how cultural bias surfaces in training data, and how interface design foregrounds certain logics while obscuring others. This method draws on theories of language, semiotics, and critical interface design to guide students through comparative mappings between AI-generated outputs and traditional design structures (e.g., grids, typographic rhythm, narrative sequencing).
Crucially, students are not passive recipients of AI assistance. They use AI as a reflective instrument to reframe and critique existing design workflows—extracting embedded design grammars, stress-testing stylistic assumptions, and making strategic use of the model’s generative excess. Students build speculative tools, experimental interfaces, and annotated systems that visualize not just results but the underlying decision tree behind them.
Through this process, students achieve more than tool fluency—they cultivate a critical authorship grounded in system thinking, capable of navigating the noise of generative output with informed judgment. The outcomes show that once AI is treated as an epistemological partner rather than a shortcut, students are empowered to articulate design decisions with greater clarity, ethics, and intentionality.
This design research is presented at Design Incubation Colloquium 12.2: Annual CAA Conference 2026 (In-person only) on Thursday, February 19, 2026.