The Problem
Design education cultivates something fragile. AI is fluent enough to damage it.
The capacities that design education develops in students — perception, situated judgment, tolerance for ambiguity, material thinking, the ability to hold a problem open long enough for a genuine response to emerge — are not efficiencies. They are not skills that can be listed on a syllabus and ticked off. They are orientations. They develop slowly, through repeated encounters with difficulty, confusion, and the resistance of materials and ideas that refuse to behave as planned.
AI disrupts these capacities not by being wrong but by being fluent. It offers resolution where pedagogically productive struggle is what matters. It generates output where the student’s encounter with the process of making — the dead ends, the ugly iterations, the moment where something almost works and you cannot yet say why — is where learning actually lives.
Across Indian design institutions — NID, Srishti, CEPT, IIT, Anant, Karnavati, and the growing network of private universities — the response has been institutional, not pedagogical. Policies specify what is permitted and what is penalised. Faculty meetings produce guidelines. Academic integrity committees draft detection protocols. These are administrative responses to a pedagogical crisis. They address the symptom (students using AI) without engaging the question (what does it mean for a student’s development when the struggle that produces understanding can be bypassed?).
The landscape is fractured. One institution prohibits AI in studio submissions. Another requires it. Most occupy an uncomfortable silence between the two, where faculty use AI privately and restrict it publicly, and students use it regardless and learn to hide it. None of these positions emerge from a serious engagement with what design education actually does and what AI actually changes about it.
What Design Education Does
The design studio is not a classroom with better lighting. It is a distinct pedagogical environment — closer to an atelier, a workshop, a rehearsal room — where students learn by making things that do not yet work and discovering, through revision and critique, why they do not work. The crit is not assessment delivered from above. It is a dialogic encounter where a student defends choices they may not fully understand yet, and in the act of defending them, begins to understand. The desk review is not supervision. It is a conversation where a teacher looks at unfinished work and asks questions that help the student see what they cannot yet see on their own.
This process is slow by design. It requires confusion. A student who arrives at a resolved form without having lived through the irresolution has not learned to design — they have learned to produce. The difference matters. The designer who has sat with ambiguity, who has felt the wrongness of an iteration before they could articulate it, who has revised not because they were told to but because something in the work demanded it — that designer possesses a kind of intelligence that no output can demonstrate. It is tacit, embodied, hard-won. It lives in the hand that knows when a model’s proportion is off, in the eye that registers when a layout breathes and when it suffocates.
This is what is at stake. Not tool adoption. Not academic integrity. Not whether students should be “allowed” to use AI. What is at stake is whether the encounter with difficulty — the engine of design learning — survives when a fluent machine offers to bypass it.
Five Questions
Before any institution adopts, restricts, or ignores AI, it owes its students honest engagement with these questions:
-
What is AI actually doing in the design process? Not “is AI helpful?” but what kind of operation is being performed — pattern recognition, generation, retrieval, interpolation — and where in the student’s process is it operating? A tool that generates form options and a tool that evaluates concept coherence are performing fundamentally different operations with fundamentally different pedagogical consequences.
-
What happens to the student? When AI enters a student’s design process, what cognitive and creative capacities are being developed, and which are being bypassed? Does the student still learn to perceive, to tolerate ambiguity, to revise from felt wrongness rather than explicit instruction? Or does AI compress the very encounters — with confusion, with resistance, with not-knowing — through which those capacities form?
-
What happens to the studio? The studio is a social space — crits, desk reviews, peer learning, shared struggle. How does AI change the crit when some of the iterations on the wall were generated in minutes rather than wrestled into being over weeks? How does it change peer learning when the gap between students is no longer effort and skill but access and fluency with AI tools? What happens to the culture of the studio when the shared experience of difficulty — which bonds a cohort and builds mutual respect — is unevenly distributed?
-
What are we assessing? If students use AI in their process, what exactly is being evaluated — the student’s growth, the quality of the output, the sophistication of the AI direction, or the ability to integrate AI into a design process? Each answer implies a different pedagogy, a different kind of portfolio, a different relationship between process documentation and final work. Design assessment has always been interpretive and dialogic. Does AI make it more so, or does it push institutions toward metrics that betray the nature of design learning?
-
Who is served? Does the AI tool serve the student’s learning, the institution’s efficiency, or the platform’s growth? When a university adopts an AI-powered assessment tool, whose interest drove that decision? When a student uses a generative tool, whose convenience is being served — the student who wants to learn, or the student who wants to finish? These interests diverge more often than they align.
These are not rhetorical questions. They are the minimum conditions for a pedagogical response. Any institution that cannot engage with them has a policy, not a position.
The Roundtable’s Position
The roundtable holds three commitments:
Design education has always known this, even when assessment systems have struggled to honour it. The encounter with difficulty — with ambiguity, with materials that resist, with ideas that refuse to resolve — is not an obstacle to learning. It is the mechanism of learning. If AI accelerates output but shortcuts the encounter with confusion and revision, education has lost its purpose regardless of how polished the portfolio looks. Process is not a means to an end. In design education, process is the end.
Tacit knowledge, embodied practice, material resistance, situated judgment, the eye that knows before the mind can explain — these are not inefficiencies to be optimised away. They are how designers learn to think. They are also, precisely, the capacities that AI cannot develop in a student and that fluent AI output most effectively disguises the absence of. Protecting these ways of knowing does not mean refusing technology. It means understanding, with rigour, which pedagogical encounters cultivate them and which bypass them.
Not policy. Not technology adoption roadmaps. Not institutional efficiency metrics. The question — always, in every decision about AI in a design programme — is: what does the student learn from this? If the answer is “we don’t know yet,” that is an honest and useful answer. If the answer is never asked, no policy document can compensate.
Contributions
The roundtable gathers demonstrated approaches — not proposals, not position papers, but things people have actually tried in studios, classrooms, and tool-building practice. Each contribution engages with one or more of the five questions through work, not theory alone.
Koher Architecture
Koher contributes an architectural approach to the first question — what is AI actually doing? — through a three-layer separation of language understanding, deterministic judgment, and narration in tool design. Demonstrated through the Coherence Diagnostic, a free tool for evaluating design concept statements, where AI reads language patterns, code makes evaluative judgments, and AI narrates the result. Open source under MIT licence.
Open invitations
The roundtable invites contributions from anyone grappling with these questions in practice. Some of the questions we hear teachers asking:
- How should a design crit work when the student used AI for half the iterations?
- What does a hand-drawing assignment mean when AI can generate renderings from a text prompt?
- How do you assess a student’s design judgment when the output is AI-assisted and the process is partially invisible?
- What ethical frameworks help students see AI tools as designed objects — with embedded values, biases, and commercial interests — rather than neutral utilities?
- What happens to material exploration when students can “test” materials virtually before touching them?
If you have tried something — a studio method, an assessment approach, a curricular experiment, a tool — that engages with any of these, the roundtable wants to hear about it. Experiments that reveal limits are as valuable as experiments that succeed.
What This Means
For the studio
The studio’s pedagogy rests on making — on the resistance of materials, the feedback of the hand, the slow emergence of form through iteration. AI changes what can be made and how quickly, but the pedagogical question is not speed. It is whether the student’s relationship to the work remains generative. A student who has wrestled a concept through twelve revisions understands something about that concept that a student who generated twelve variations in an afternoon does not. The studio needs to articulate what that something is — and protect the conditions that produce it.
For curriculum
Design programmes need to teach students to see AI clearly — not as a magic box, not as a threat, but as a designed object with specific capabilities, specific limitations, and specific embedded assumptions. This means going beyond “how to use AI tools” toward understanding what AI does well (pattern recognition across language, interpolation within learned distributions) and what it cannot do (perceive, judge in context, feel the wrongness of a proportion, sit with ambiguity). Students who understand this distinction can make informed decisions about where AI belongs in their process and where it does not.
For assessment
Design assessment has always been interpretive, contextual, and dialogic. A jury does not score a project on a rubric and leave. It engages with the student’s intent, process, choices, and growth. AI-assisted work does not break this model — but it demands that the dialogue deepen. The portfolio must show the student’s journey, not just their destination. Process documentation becomes not a bureaucratic requirement but the primary evidence of learning. The question “what did you learn?” becomes more important than “what did you make?” — which, in good design education, it always was.
For institutions
The move that matters is from policy to pedagogy. An AI policy tells students what they may and may not do. A pedagogical position tells them why — what design education is trying to cultivate in them, what encounters with difficulty are necessary for that cultivation, and where AI helps or hinders those encounters. Prohibition without understanding breeds resentment and evasion. Permission without understanding breeds dependence. Understanding — shared between faculty and students — is the only foundation that holds.
The Roundtable
A gathering of design educators and students across Indian institutions who recognise that the field’s response to AI needs to start from pedagogy, not policy. Teachers talking to other teachers about what they are seeing in their studios, their classrooms, and their students’ work.
What it is
- Faculty and students across Indian design institutions, in conversation
- A shared commitment to understanding what AI changes about how students learn to design
- Practice-grounded: members contribute approaches they have actually tried — in studios, in curricula, in tools
- Open to multiple approaches — the five questions are shared; the answers need not be
What it is not
- An institutional consortium — no institution is a member; people are
- A standards body — no certification, no compliance, no enforcement
- A policy group — we are interested in pedagogy, not administration
- A closed circle — this conversation is open; the position document is open for co-authorship
How participation works
Participants are individuals. You might teach at NID Ahmedabad or NID Bangalore, at Srishti or CEPT, at IIT Bombay or IIT Hyderabad, at Anant, Karnavati, or one of the many programmes where design is taught under other names. Your institutional affiliation is context, not credential. No endorsement is needed.
Participation is practice:
- Contribute something you have tried — a studio method, a curricular experiment, a tool, an assessment approach — that engages with any of the five questions
- Critique existing contributions — where do they hold? where do they break? what do they miss?
- Observe what happens when AI enters your studio or classroom — and share what you see, even when it complicates your own position
- Ask the questions your students are already asking you, the ones you do not yet have good answers for