Agent-based cultural evolution / naming · myth · taboo

The Lexicon
Observatory

A small population of agents starts with no shared words, no stories, no rules. Watch a vocabulary, a drifting myth, and a taboo assemble themselves out of nothing but who talks to whom — and who gets copied. The selection pressure here isn't a fitness function. It's social transmission.

naming game · coordination myth · telephone drift taboo · norm enforcement seed 12345

Run

Auto-pause at consensus halt once ≥95% agree

World

Transmission bias

Prestige bias high-status agents copied more
Conformist bias copy the local majority

Cultural layers

Myth transmission narrative passed & mutated
Taboo norms prohibitions spread as norms
Enforcement violators lose prestige
Metanorm pressure on non-punishers

Seed

structural change — press Reset to apply

Analysis

Space run / pause · N new world · S step · R reset
hover the field to inspect any agent

The agent field — colour = each agent's current word for

Tick0
Convergence0%
Entropy0.00 bits
idle · press Run
each dot is an agent · same colour = same word
01

Live instrumentation

A simulation with no metrics produces anecdotes, not results. Every mechanism here ships with the measure that detects whether it is doing anything. Charts update as the run advances.

Lexical convergence
0%
share of population on the dominant word · target ≥ 80%
Naming entropy
0.00 bits
avg uncertainty per referent · falls as words die out
Word survival
0 / 0
alive vs. ever coined · the rest went extinct
Myth divergence
0.0
edit distance from the origin story · the telephone effect
Taboo prevalence
0%
share holding the focal taboo · time-to-fixation
Prestige → influence
r = 0.00
does status predict cultural reach? each dot = one agent
02

Does the pattern hold?

A single run is an anecdote. This sweeps the current configuration across many seeds — running each headlessly to a fixed horizon — and shows the spread of outcomes. Change a control, re-sweep, and watch the distribution move. This is the honesty check section 04 asks for, built in.

→ 400 ticks each
Final lexical convergence · distribution across seeds
status
not run yet
press Run sweep
why it matters
one seed lies
sweep before trusting a pattern
03

Inside the culture

The metrics above are summaries. Here is the actual state the agents hold — the competing words, the drifted story, the spreading prohibition — read straight from the population.

Origin story (seed)
Dominant variant now

Highlighted motifs differ from the origin. Drift accumulates through omission, embellishment, substitution and recombination on each retelling.

Surviving variants
Focal prevalence
0%
Mean register size
0.0
Status of focal
Prohibition strength across behaviours
04

What this model does & does not claim

This artifact is built to be measured, not narrated. The honesty guardrails below are part of the deliverable, not a disclaimer bolted on.

The neuroevolution analogy stated precisely

This reuses evolving-population machinery — variation, copying, differential propagation — but the objective fitness function is replaced by social transmission: a trait spreads because it is imitated, trusted, or enforced.

  • Where it holds: population, variation, selective copying.
  • Where it breaks: there is no global objective and no guaranteed convergence to "better" — only to more transmissible. A dominant word or myth is the most-copied one, not the best one.

Emergence is not guaranteed failure modes

Depending on parameters, the same engine can land in very different regimes. None of these is a bug:

  • Consensus — well-mixed naming reliably converges to one word per referent.
  • Fragmentation — spatial topology holds many regional dialects indefinitely; low global convergence here means dialects, not failure.
  • Drift without canon — myths often keep mutating without any single version winning.

Sources verify the primaries

The mechanics draw on real research traditions, but I am not citing specific papers I cannot verify from memory, and you should confirm primaries before quoting them:

  • The naming game / language-coordination family is associated with Luc Steels and collaborators (semiotic dynamics). I'm confident the broad model exists; verify specific papers/results.
  • Norm and metanorm dynamics are associated with Robert Axelrod's work on the evolution of norms. I have not verified an exact title/year here — treat the implementation as inspired by, not a reproduction of, a named model.

Reproducibility & scope seeded

Every stochastic choice is drawn from a single seeded PRNG (mulberry32) threaded through the run, so seed + parameters reproduce a trajectory exactly (verified bit-identical in testing).

  • Per-tick summary statistics are logged; full per-agent history is not — a deliberate budget choice for in-browser scale. Treat any single run as one sample; sweep seeds before trusting a pattern.
  • "Myths" and "taboos" here are data structures with transmission rules, not beliefs or fears.
Determinism
Bit-identical
same seed → same run (verified)
Naming baseline
10 / 10 seeds
reached ≥0.8 convergence in tests
Prestige bias
r: 0.16 → 0.83
off vs on, seed 4 (approx.)
Spatial topology
~73 words
persist as dialects (seed 3, approx.)