Closed loops converge
In our own tests, ideas that are revised together, by one model or a group of model roles, slide back toward the typical answer round after round. The only arm that kept ideas apart brought new outside material into every round. Random fragments kept them far but made them less credible; real facts kept them far and credible (Closed loops converge).
Our hypothesis is that human invention escapes collapse the same way. It is not a different generator. It is an open loop of fresh evidence and selection.
What the research on living abroad says
The best natural experiment for outside material is moving to another country. Studies of it agree on a few points.
- Living there counts; visiting does not. Time spent living abroad predicts creative performance, while time spent travelling does not, and the effect runs through adapting to the new culture (Maddux and Galinsky, 2009).
- Depth matters most. In 270 fashion houses, creative directors’ work abroad raised the creativity of their collections. Depth of experience had the largest effect; breadth across countries and cultural distance helped most at moderate levels (Godart et al., 2015).
- The turn often has a date. The Korean painter Kim Whanki’s move to abstraction followed the 1963 São Paulo Biennale and the New York years that came after it.
A natural experiment in our own data
Our main framework study follows the Spanish architect Alberto Campo Baeza, with his work split at 2008 (Sounding like him is not predicting him). In 2009 he visited Luis Barragán’s houses in Mexico for the first time, just after our cutoff. If outside material changes a person’s distribution of ideas, his later work should show it, and a framework written only from his life before 2009 should miss it.
What we propose
Two changes follow.
- Dated foreign-input records. A person’s framework should record when and where they took in something foreign, so a model can tell which of their moves came from outside.
- A residency step in the workflow. Give the generator deep, real material from a moderately distant place, ask it to solve a local brief there first, then return home with a list of what differs from the person’s own axioms.
The first decides where on a person’s curve to pick an idea. The second is about moving the curve itself, which is closer to what a breakthrough is.
Sources
- William W. Maddux and Adam D. Galinsky, Cultural borders and mental barriers: the relationship between living abroad and creativity, Journal of Personality and Social Psychology, 2009.
- Frédéric C. Godart, William W. Maddux, Andrew V. Shipilov and Adam D. Galinsky, Fashion with a foreign flair, Academy of Management Journal, 2015.
- Feathers & Wax, flux and evidence runs, code in
research/cllm-divergence.