← Read the essay

Where the argument stands

A status-tagged audit of the cultural-selection research. Every evidenced claim points to an observation; partial claims expose their limits; open claims name what the current experiments cannot decide.

✓ evidenced◐ partial○ openTODO
C1

Culture causally changes what otherwise identical makers produce. ◐ partial

C1.1

A blind classifier assigned 36 of 36 v2 artworks to the correct culture. ✓ evidenced

Evidence: LOG.md (2026-08-06 21:10); keyed blind set recorded in results/art_assets.json; chance expectation 6/36

C1.2

Charter identity produced between/within feature ratios of 449 for drawable elements, 212 for bytes and 90 for colour count in v2. ✓ evidenced

Evidence: v2 fleet; 18 worlds

C1.3

An earlier shared-seed fleet retained the same circular composition across different creeds, showing that visual ancestry constrains doctrine. ◐ partial

Evidence: v1 versus v2 structural comparison; n_drawable F-ratio 0.74 versus 449

C2

Coherent culture-level styles can arise from distributed local choices without a central designer. ◐ partial

C2.1

Each world accumulated proposals and adoptions from 24 agents; no role selected a target final style. ✓ evidenced

Evidence: simulation rules and accepted-version histories

C2.2

Matched Open Circuit lineages visibly diverged inside ascetic, baroque and charter-free worlds. ✓ evidenced

Evidence: interactive trace reader; six accepted snapshots per world

C2.3

The culture-level trajectory was unplanned even when individual revisions cited explicit aesthetic reasons. ◐ partial

Evidence: accepted-version reasons in outputs/2026-08-06-twin-worlds-v2

C3

Some artistic guidance is transmitted without being explicitly represented by the maker. ○ open

C3.1

The present design bundles a written charter, visible ancestor and adoption history, so their separate effects are not identified. ✓ evidenced

Evidence: v2 design and preregistration

C3.2

The fraction of artistic divergence caused by explicit doctrine versus visual inheritance or tacit convention is unknown. ○ open

Evidence: Requires factorial ablation; traces alone cannot identify it

C3.3

Cross charter presence, visual ancestor and access to lineage history independently. TODO

Evidence: Proposed 2×2×2 causal design

C4

Cultures may create different balances between conformity and expressive range. ◐ partial

C4.1

Charter identity accounts descriptively for 85.3% of accepted-depth and visual-dispersion variance across 12 worlds. ◐ partial

Evidence: results/depth_breadth_summary.md

C4.2

Cave and psychedelic worlds separate at the extremes while four middle cultures overlap. ◐ partial

Evidence: two replicates per culture; cave-r1 is sparse

C4.3

Hold aesthetics fixed and manipulate conformity rewards to test whether the social rule causes expressive range. TODO

Evidence: Proposed fleet

C5

Hierarchy and population structure may generate distinct artistic traditions. TODO

C5.1

Compare centralized patronage with distributed proposal and adoption. TODO

Evidence: Proposed hierarchy intervention

C5.2

Compare isolated native populations with migration and cultural mixing. TODO

Evidence: Proposed population intervention

C5.3

Human viewers recognise, value and explain the resulting differences. ○ open

Evidence: Requires blinded human study

Unresolved / TODO

  1. Factor verbal charter, visual ancestor and lineage-history access independently to identify explicit versus inherited guidance.
  2. Run the social-structure fleet: conformity versus novelty, centralized patronage versus distributed adoption, isolation versus migration.
  3. Add replicates before treating conformity/expressivity regimes as inferential rather than descriptive.
  4. Run a blinded human recognisability study, followed by interviews about which visual cues people used.