1a. Schelling Segregation Model

Each agent on a grid wants at least a certain percentage of its neighbors to be of the same type. If unhappy, the agent moves to a random empty cell. Even mild preferences lead to dramatic segregation.

Controls

Step0
Happy0%
Unhappy0
Segregation Index0.00
Moves0

Mechanism

Even mild individual preferences (as low as 33% same-type neighbors) lead to extreme macro-level segregation. This is emergence: micro motives create macro behavior. No individual wants segregation, yet the collective dynamics produce it. Each move by an unhappy agent changes the neighborhood composition for others, creating cascading relocations.

Real-World Examples

  • Residential segregation in American cities -- racial neighborhoods persist even as individual prejudice declines
  • Self-sorting in school cafeterias -- students cluster by interest/identity without explicit rules
  • Political echo chambers online -- mild preference for like-minded content produces extreme filter bubbles
  • Gentrification dynamics -- economic thresholds drive neighborhood-level transformation

Key Insight

Tolerance does not guarantee integration. Schelling showed that even tolerant individuals contribute to segregation when personal thresholds for minimum group representation are crossed. The system-level outcome (segregation) is far more extreme than any individual's preference would suggest.

1b. Granovetter's Threshold Model

Each person has a threshold: the number of others who must act before they will join. A cascade begins when the first movers activate those with low thresholds, who in turn trigger those with higher thresholds.

Controls

Round0
Active0
% Active0%
StatusReady

Mechanism

Small changes in the threshold distribution can cause or prevent full cascades. If a uniform distribution has thresholds 0,1,2,...N-1, a complete cascade occurs. But removing just the person with threshold=1 breaks the chain entirely. The diversity of thresholds matters as much as the average threshold. This is a tipping point model -- gradual input produces sudden output.

Real-World Examples

  • Protests and riots -- a few instigators can trigger mass participation if threshold distribution is right
  • Technology adoption -- early adopters trigger mainstream users (Rogers diffusion curve)
  • Fashion trends -- trendsetters lower the threshold for others to follow
  • Bank runs and financial panics -- each withdrawal makes the next more likely

Key Insight

Two populations with the same average threshold can produce completely different outcomes. What matters is the full distribution -- especially whether there are enough low-threshold individuals to start the chain. Granovetter showed that predicting collective action requires knowing the entire distribution, not just the average preference.

1c. Standing Ovation Model

An audience decides whether to stand based on their perception of quality plus social pressure from neighbors already standing. Higher quality triggers more initial standers, which cascades through the audience.

Controls

Round0
Standing0
% Standing0%
StatusReady

Mechanism

Each audience member receives a noisy signal of quality. If the signal exceeds their personal standing threshold, they stand immediately. Then, social pressure kicks in: seeing neighbors stand lowers the threshold for others. Higher quality causes more initial standers, which creates a cascade. Audience diversity (variation in thresholds) and social connectivity both matter for whether partial or full ovation occurs.

Real-World Examples

  • Actual standing ovations -- quality matters, but so does seating arrangement and social norms
  • Viral content -- initial engagement triggers algorithmic amplification (social pressure equivalent)
  • Product reviews -- early positive reviews cascade into more positive reviews
  • Restaurant popularity -- visible crowds attract more diners (information cascades)

Key Insight

The same performance can receive a standing ovation one night and polite applause the next. Randomness in initial signals, combined with social amplification, means outcomes are path-dependent. This explains why "going viral" is so hard to predict -- it depends on the early stochastic response as much as the underlying quality.