Methodology
1. Value Extraction:
For each essay (location), extract (dimension, value) pairs from the best hypothesis.
Values are truncated to 100 chars for comparison.
2. Shared Value Computation:
For each pair of essays (A, B), count how many (dimension, value) pairs they share.
shared(A, B) = |values(A) ∩ values(B)|
3. Total Connectivity:
For each essay, sum shared values across all other essays.
connectivity(A) = Σ shared(A, X) for all X ≠ A
4. Clustering:
Essays with high mutual shared values form clusters.
Core cluster: reason, memes, horus, violence, reproduction, tantra (40-45 shared pairs)
Outliers: communes, water (0 shared with anyone)
5. Late Bloomer Detection:
Compare fitness at generation ≤10 vs fitness at generation >10.
Late bloomers improved by ≥30% after early generations.
6. Bridge Essay Identification:
For late bloomers, find essays with highest shared value count.
These bridges may have contributed successful dimension values via crossover.