TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · Judgment · AGE 17—22 · POSITIONAL · HELD
Taste is trained
The machine's taste is the average of everyone's, and yours has to be the opposite.
Our position
Taste is not a gift. It is a reference library you built, a set of positions you hold, and a growing sense of what is yours and what is noise. You build it on purpose: by looking at more than you are shown, by keeping what earned its place and cutting what did not, by being able to say why — and by asking who chose what you were shown. It is the one filter the machine cannot give you, because the machine's taste is the average of everyone's, and yours has to be the opposite of that.
Why we hold it
Generated output is plentiful and arrives already looking good. The only thing that sorts it is a filter the person brings, and a filter is only as good as what it was trained on. A student with twenty references they can defend reads a batch differently from a student with a feed. The Lab teaches taste as a discipline (a library, positions, deletions) because the alternative is to inherit the model's.
The strongest objection
Taste is not an individual library built "on purpose." What a student meets as good taste has already been filtered — by schools, publishers, awards, class, caste, wealth, colonial archives, the algorithms of the feed, and by who had the money to be photographed and remembered. A curriculum that says "train your taste" without saying so is teaching students to internalise the canon more efficiently, and to mistake the canon's filter for their own. In India that filter has left out most of what was built. And if generative tools raise everyone's floor — the research says they can, most of all for the less practised — the market may stop paying for a personal filter at all.
What would make us revise it
The first half of the objection we take as a correction to the card, not a refutation of it: a library built on purpose must include an account of who filtered what reached you, and the deletions must include inherited ones. We would revise the position itself on evidence from practice that offices stopped distinguishing, in hiring, in fees, and in who gets the room, between a trained filter and a good tool. If the market stops paying for the difference, the card moves from "the job" to "the craft." That is a real demotion and we would make it honestly. We watch practice, not only studios.
Try it
Build a twenty-item reference library (buildings, drawings, photographs, texts) and write one line on why each item earned its place. Then state one position you hold that would change how you brief a model. "I distrust roofs that do not show their thickness," for instance, is a position, and it writes a line of every brief.
Take it to crit
Ask what the student has cut from their references this term, and why. Taste shows in what gets cut. A library that only grows is just a feed in a folder.
How it works
Three parts, and each one can be built. The library: things you have looked at long enough to say why they matter, which is a different act from saving them. The positions: the few rules you hold that survive tool changes, and that can be written into a brief. The deletions: what you once kept and no longer do, which is the clearest record that the filter is moving. The Lab's own judgment register, which is internal material and not public, is built the same way, and this card is its student-facing form.
What this idea builds on
What this idea opens up
Sources
- judgment_register
- Doshi & Hauser, Science Advances 2024
- The Seduction of the Machine
Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.