TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · Judgment · AGE 18—21 · FACTUAL · EVOLVING
Individual lift, collective sameness
Each student is a little better, and the room is a lot more alike.
The dilemma
The tool seems to make your work better: a longer idea list, sharper images. The same seems true for everyone else in the studio, on the same tool. Walk the pin-up at the end of term and ask whether the room is richer or narrower than it was last year.
The choices
Judge the tool by what it does for you. Or judge it at two scales at once: what it does for one person, and what it does to the variety across the group. The two can move in opposite directions.
The consequence
Judge only the individual lift and a studio of thirty drifts towards one model's habits: the same roof, the same light, the same courtyard. Each student is a little better and the room is a lot more alike. Judge both scales and you can take the lift and protect the variety, with different models, different briefs, images introduced late, and precedent drawn from the real world rather than from the same feed.
The case
Picture an end-of-term exhibition in the first year a studio uses image tools widely. A visiting juror says the work is more polished than last year's. A minute later she says she cannot tell the projects apart. Both observations can be true at once. No single student did anything wrong. The room as a whole did.
Try it
Collect a class's AI-assisted outputs for one brief and measure how similar they became: sort them into families and count. Then explain, in two sentences, how a tool can help you individually and make the group more alike at the same time.
Take it to crit
Can the student show where the class converged on one model's taste, and say what it would take to break out? Ask them which of their own moves they are confident nobody else in the room made.
How it works
Two studies put numbers on the pattern. In one, writers given AI story ideas produced work that was rated better, and the lift was largest for the less practised writers, but the stories across the group were more similar to each other than the ones written unaided. In another, people generating ideas with a language model converged on a narrower set of ideas than people working without one. Both studies were on writing and idea lists, not on architecture studios. The studio drift described here is the Lab's inference from them and from what it has seen at pin-ups, and it should be read as that. EVOLVING, because the pull towards sameness depends on everyone using the same few models, and on interface choices that could change.
What this idea builds on
What this idea opens up
- Nothing yet names this as a foundation.
Sources
Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.