Arjun Acharya

Collapse of societies - will trust survive an anti-trust society from an evolutionary lens.

Arjun Acharya
7 min read · Feb 6, 2026

I’m modeling how distrust spreads faster than trust. Simple rules. Clear asymmetry. One bad move can outweigh a lot of good. Elites amplify both. We run this on two grids and watch what emerges.

Rules of the world and how individuals interract with each other.

In our simulation, every person has a trust score. Good actions raise it, bad actions lower it. Trust is unbounded—it can grow infinitely positive or fall infinitely negative, creating a continuous spectrum of behavior.

Who’s in the world?

  • The Powerful – the ones with the most influence.
  • The Common – everyday people who make up much of the middle.
  • The Poor – the most vulnerable group.

Players & ratios

Powerful (10%), Common (30%), Poor (60%). Ratios stay steady—we’re focused on trust flow, not class mobility.

How they affect each other

People interact with their 8 closest neighbours. Each interaction either builds trust or damages it, but not equally – bad actions usually have a stronger impact than good ones.

  • The Powerful’s actions ripple out the most – they can lift trust a lot or damage it badly.
  • Common folk influence each other and the poor more than they influence the Powerful.
  • The Poor have the weakest reach, but they still shift trust in small ways.

The two worlds

We start with two halves:Good World, where everyone begins with maximum trust (+10), and Mixed World, where trust levels are randomly spread between good and bad.

Migration

If someone in the Mixed World experiences negative impacts from their neighbors for four consecutive ticks (meaning their environment is consistently harmful), they "migrate" to the Good World – but not as a good person. They replace a random good person there, bringing their distrust into the good side and leaving their old spot empty.

Why it matters

Over time, we can watch how distrust spreads like an infection, and how a few bad actors can slowly unravel even the most trusting societies.

What you’ll see

Two worlds side‑by‑side: Good (everyone starts at +10) and Mixed (good + evil). Each cell listens to its 8 neighbors and adjusts. Colors: green = high trust, yellow = neutral edge, red = entrenched distrust. Histograms below track how the population shifts over time.

Trust & moves

  • Trust score: (−∞, +∞). Unbounded in both directions—trust can grow infinitely or collapse completely.
  • Move rule: Uses a logistic function: P(good) = 1 / (1 + e^(-α·T)), where α = 0.25. This creates a smooth S-curve: very negative trust → ~0% good actions, T=0 → 50% good, very positive trust → ~100% good.
  • Neighborhood: 8 neighbors (Moore). Your update is the average impact of their moves on you.
  • Self‑drift (Powerful only): +0.1 if they act good, −0.1 if they act evil.

Scoring (impact) — actor → target per interaction

Normalized deltas. Bad hits harder than good helps. We apply these neighbor‑wise, then average.

ActorTargetGood ΔEvil ΔComment
PowerfulAnyone+0.5−1.0Megaphone effect
CommonPowerful+0.1−0.1Upward impact is small
CommonCommon+0.1−0.5In‑group betrayal hurts
CommonPoor+0.2−0.5Downward harm is heavier
PoorPowerful+0.1−0.1Barely registers upward
PoorCommon+0.1−0.2Moderate
PoorPoor+0.1−0.3Local harm spreads fastest

Worlds & migration

  • Good world: everyone starts at T = 10.
  • Mixed world: random in [−10, +10] at start; voids can appear.
  • Migration (one‑way): in Mixed, if an actor experiences negative net impact from neighbors (delta < 0) for 4 consecutive ticks, they migrate to overwrite a random Good cell; the Mixed slot becomes a void.
  • Randomization: optional shuffle of positions each tick.

Reading the heatmap

Green → good, yellow → starting to distrust, red → entrenched distrust.

50%
12 tps
0.25
α (alpha) controls how trust affects behavior. Lower α = gradual change (trust must build high for consistent good actions). Higher α = sharp transitions (small trust changes flip behavior quickly).
tick 0 · queued 0 · migrations 0 · meanT(G) 0.00 · meanT(M) 0.00 · G(g/e) 0/0 · M(g/e) 0/0

Good World

Mixed World

Good World – Distribution

Distribution by trust

Mixed World – Distribution

Distribution by trust