civ_sim · v7.4
Experimental toy model · cultural memory · 700 simulated years

Cultural Memory
Toy Simulator v7.4 — claim reasons

Objects become sacred, taboo, or contested through remembered events and social feedback, rather than because anyone explicitly scripted them that way.

Not a historical model — a small causal sandbox.
01

The Question This Model Asks

The model asks how an object stops being merely a tool. A bowl, cup, bone fragment, or scrap of writing may begin as an ordinary thing in everyday life. But once events are remembered, retold, and attached to places or rituals, the object can become sacred, avoided, or contested by groups that disagree over its meaning or custody.

Question 1

When does an object stop being just an object? The model follows how everyday goods and relics acquire significance through memory and repetition.

Question 2

How do records and stories drift apart? Practical records stay close to events, while narrative and mythic records compress, emphasize, and begin to mean something else.

Question 3

Why do shared objects become disputed? Even when groups share the same relic, differences in origin, custody, ritual, or interpretation can turn it from a bond into a source of conflict.

Question 4

Is myth a cause, an effect, or part of a loop? The model does not treat myth as a universal cause. Myth moves as one part of a feedback loop involving objects, places, records, and action.

In that sense, this is a small experiment in reading culture not only as belief, but as a relation among objects, places, records, and repeated action.

02

What it simulates

Four tribes live for about seven hundred simulated years. Each one observes daily life, turns those observations into memory traces, and slowly builds attitudes toward the objects and places around it. The physical layer — food, population, war, trade goods — is kept deliberately simple so the cultural layer on top of it is the thing under observation.

River

fish · reeds · clay

Lives by water and the harbor; its everyday objects are nets, jars, and boat pegs.

Hill

stone · copper · goats

Quarries and herds; carries seals, antlers, and metal tokens.

Forest

herbs · timber · honey

Gathers and hunts; keeps charms, masks, and herb pouches.

Salt

salt · dried meat · dye

Trades and tallies; holds jars, weights, and ledger shards.

Objects live in three tiers
local — a tribe's own everyday material culture regional — relics shared between neighbours universal — broadly shared, religion / kingship scale

Crucially, the model keeps separate records of the same events. A practical record stays close to what actually happened; a narrative record and a mythic record are free to drift. Watching the gap open between them is the point — the drift is observed, not asserted.

practical
narrative
myth

Same starting event — three records of it, allowed to diverge over time.

03

The core loop

Everything important runs through one closed cycle. Tribes observe life, turn those observations into memory traces, condense recurring patterns into symbols, let symbols bias their attitude toward specific places, and let that bias reweight what they do day to day — which produces new observations. The loop, not any single rule, is what the project exists to test.

observation daily life is watched trace memory sediment_memory() symbol recurring traces condense place bias attitude toward places daily life behaviour reweighted claim_reason read-only · v7.4
v7.4 ↗

The new claim_reason layer sits beside the loop, attached by a dashed line. It records why a contested object is disputed, but it is never read by food, war, symbol, or place-bias logic. Removing it leaves every simulation trajectory unchanged — it explains the loop without entering it.

04

Claim reasons in v7.4

When two tribes hold conflicting attitudes toward a shared relic, the object becomes contested rather than simply flipped to evil. v7.4 attaches a reason to that dispute — six categories that read like plausible cultural friction. They are labels, not levers.

ownership

Whose object is it? Competing claims of custodianship over a thing held in common.

origin_claim

Each group claims a more authentic origin for the relic.

ritual_method

Neither side accepts the other's rite or handling of it.

purity

The other side's handling is judged impure.

custody

Neither tribe will surrender the keeping of the object.

translation

The text or meaning is read differently by each side.

Which reasons each tier leans toward (weighted, not strict)
PoolLeans toward
local objectsownership · custody
regional relicsorigin_claim · custody · ritual_method
universal relicsorigin_claim · translation · purity
Book of the Deadtranslation · origin_claim · purity

Everyday objects are fought over as property; broadly shared relics are fought over as interpretation. The tendency follows the tier, which is what makes the disputes legible rather than random.

Because the reason comes from a sha256 of tribe · object · pool, the same tribe always gives the same object the same reason — reproducible across runs and machines, and without touching the global random state.

05

Ablation results

The ablation study runs the model under eight conditions, each disabling one mechanism, and reports the mean of every metric. The question for each row is simple: what changes when this mechanism is removed? The numbers below are a behaviour check inside this toy model — not proof of anything outside it.

The spine

Place bias drops to zero

NO_OBSERVATION_SEDIMENT · NO_SYMBOL_FEEDBACK
0.24 → 0.00

Cut observation sediment and no symbols form at all; cut only the feedback and symbols still form (44) but stop biasing places. The symbol → place-bias channel is load-bearing.

Object conflict

Disputes empty out

NO_MYTH · NO_OBJECTS
17.7 → 0

With no myth or no objects, contested_objects and trade_refusals both fall to zero. Myth is the engine that turns events into disputed relics.

Worldview → action

Trust and diversity slip

NO_WORLDVIEW_TO_ACTION
4.25 → 3.19

Trust, worldview diversity, sacred places and contested objects all move down together — the worldview→action path drives the social metrics.

Selected metrics · mean per run · N=30 seeds · 700 years
metric CONTROL NO_OBS_SED NO_SYM_FB NO_WV→ACT NO_MYTH NO_OBJECTS
symbol_count44.000.0044.0044.0044.0044.00
avg_abs_place_bias0.240.000.000.250.240.25
contested_objects17.6716.0017.678.630.000.00
trade_refusals17.8016.2717.9315.470.000.00
avg_trust4.254.204.193.194.674.67
worldview_diversity2.572.802.801.772.572.63
myths1304.501298.931292.831215.830.001288.13
wars20.6323.2021.6721.609.9020.37
hard_relic_reversals0.000.000.000.000.000.00

A few things are worth reading together. The contested_reason_* counts also fall to zero under NO_MYTH and NO_OBJECTS, confirming the reason layer is wired to the object machinery rather than generated on its own. And hard_relic_reversals stays at zero everywhere — the same object is never sacred to one tribe and taboo to another; conflict routes into contestation instead.

How to read these. Effect sizes (Cohen's d) and means describe how this particular model behaves when a part is removed. They show that each mechanism does measurable work inside the sandbox. They are not evidence about real cultures, and nothing here is offered as proof.

06

Why it matters

The interesting claim is small and structural. Cultural difference here is not produced by individual minds remembering things. It emerges from a loop in which everyday objects, places, rites, and refusals to trade feed back into behaviour and slowly pull two groups apart.

Myth is one mechanism among several — not a universal cause. Objects, places, the practical record and the mythologized record drift on their own tracks.

When myth is removed, the object-conflict layer empties; but local objects survive partially because they come from everyday observation, not only from the relic machinery. When objects are removed, trust rises — because relic disputes were precisely what had been suppressing it. The model lets you watch these threads pull against each other instead of collapsing every outcome into a single story.

07

Files & reproducibility

Pure Python, no third-party dependencies. The reason layer is deterministic, so seeded runs reproduce exactly across machines.

Python 3.10+ no external deps 4 tribes · ~700 years seeded · reproducible
# run one seeded narrative (prints a chronicle) python civ_sim_v7_4_claim_reasons.py # run the full ablation study (set N near the top) python ablation_v7_4_claim_reasons.py # CONTROL-only reason breakdown + refusals python aux_v7_4_control_reasons.py
  • civ_sim_v7_4_claim_reasons.py
  • the model — tribes, memory, symbols, objects, claim reasons
  • ablation_v7_4_claim_reasons.py
  • eight-condition study, mean ± SD and Cohen's d vs control
  • aux_v7_4_control_reasons.py
  • control-only: Book-of-the-Dead reasons + refusals by reason