The idea
Our second hypothesis is that the astonishing ideas sit at the far end of a person’s own curve. One way to get there might be the person’s own tensions: the places where their principles pull against each other. Campo Baeza’s framework lists 14 of them.
The test
Five arms each wrote 64 ideas across eight briefs: plain, framework only, framework plus a random seed, framework plus an inverted refusal, and framework plus one of his tensions. We measured distance from his typical framework idea, and asked a lineup judge whether each idea was his and a critic holding his framework whether he would sign it.
| Arm | Far from his typical | Picked as his | Far, picked and signed |
|---|---|---|---|
| Framework only | 8% | 91% | 3 of 64 |
| + random seed | 30% | 94% | 10 of 64 |
| + inverted refusal | 41% | 69% | 5 of 64 |
| + his tension | 61% | 77% | 10 of 64 |
| His real projects | 89% | 66% | — |
What it shows
Tension seeds reach the far band twice as often as random seeds and still read as his about as well as his real projects do. But their far ideas read as him less reliably, so both arms end with the same number of ideas that are far, recognisably his and signed: 10 of 64.
His real projects are a reminder of how far a real person ranges. Most of them sit far from the framework’s typical idea, and the judge finds only two in three of them to be his.
Limits
All judging is by models that read his framework, and the framework arm’s caricature outscores his real projects. No arm gets far from his real work on the embedding, so this cannot show an idea is new to him. Only he, or architects who know his work, can.
Sources
- Feathers & Wax, tensions pilot, code in
research/campo-baeza-tensions.