TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · Judgment · AGE 19—22 · POSITIONAL · HELD
When not to generate at all
A workflow with no refusal in it has been designed by the tool.
Our position
The sharpest judgment in this strand is knowing when the machine should stay out. Sometimes a hand sketch, a site visit, or a conversation with the person who will live there is the right tool, and generation is the wrong one. Not because it would fail, but because it would succeed at the wrong thing: filling the moment with plausible output when the moment needed attention. Refusal is a design move. A confident "no", with a reason behind it, at the right moment, and defensible afterwards, is where taste shows most.
Why we hold it
A tool that is always available is always a temptation, and the cost of generating is now low enough that the only thing stopping you is judgment. The Lab holds that a designer's workflow should contain places where the machine is excluded on purpose (the first hour with a brief, the site, the client's kitchen), and that the student should be able to say where those places are and why. A workflow with no "no" in it has been designed by the tool.
The strongest objection
Refusal can be habit dressed up as judgment. "I keep the machine out of early concept" is often a preference formed before the tools were any good, and the profession has a long record of refusing new instruments on principle and then adopting them late, in a hurry. Worse, a taught "no" becomes a ritual. Students exclude the tool at the stage they were told to, without ever testing whether that stage is where it actually fails them. Adoption data across the profession suggests the people refusing are not, on the whole, the people judging.
What would make us revise it
The position requires the "no" to be reasoned. The student must be able to say what the machine would have cost at that stage, and must have tested it at least once. We would revise the card if we found that students who excluded generation at a stage consistently produced weaker work at that stage than students who used it critically, which would mean the "no" was costing the work rather than protecting it. That is a comparison workshops can run, and the Lab runs it.
Try it
Name a stage of a real project where you kept the machine out on purpose, and defend that call. What would it have cost, and how do you know? Then take two tasks from the same project and separate the one where generation adds value from the one where it only adds noise.
Take it to crit
Ask the student where in their project they refused to generate. Was that a reasoned call, or habit? Ask whether they tested the alternative. A reasoned "no" has a comparison behind it.
How it works
Two kinds of evidence sit behind the position, and they point in slightly different directions, which is why this is a held position and not a rule. On one side, a study of AI assistance in vocational classrooms found lower-performing students offloading thinking rather than deepening it, and a 2025 retrospective, observational study in one clinical setting reported endoscopists' unassisted detection rates dropping after a period of AI-assisted work. That is early evidence of deskilling, in one field; the wider literature on offloading is mixed, and none of it is from architecture. On the other side, surveys of practice show adoption rising steadily, and the history of tools in this profession rewards early critical engagement over refusal. The card sits on the line between them. Exclude on purpose, test the exclusion, and keep the skill the machine would otherwise quietly take.
What this idea builds on
What this idea opens up
Sources
- iiid-saurashtra_ai-fundamentals
- judgment_register
- HCM.C
- Endoscopist deskilling risk after exposure to AI in colonoscopy — The Lancet Gastroenterology & Hepatology, 2025
- From Co-Design to Metacognitive Laziness
- RIBA AI Report 2025
Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.