When there is a very large personal bias, results are strange.

[docmerlin]

When they are not happy if there is a single of the other shape near them (I.e.: They are extremely shapeist), segregation slowly goes down over time.

[vassudanagunta]

I think that's a combination of the two factors:

1. When they are extremely shapeist, there simply isn't enough space on the board for them to find "satisfaction", so they keep moving endlessly.
2. They move randomly.

Because of these two factors, the static equilibrium you find on less shapeist boards is not possible, and the dynamic equlibrium tends toward zero simply due to the constant random moves, a sort of Brownian motion.

There are two ways this could be changed to better reflect reality (or what our intuitions think is reality):

• Increase the free space on the board. In which case it would quickly find static equilibrium at 100% segregation.
• Instead of moving randomly, the shapes move to the "best available neighborhood", as real shapeists would do. If this were done, then static equilibrium would be reached closer to the maximum possible segregation given the amount of free space. It would usually not reach the maximum possible segregation though, because no matter how shapeist they are local selfish moves are don't usually find gobally optimal solutions. Rather ironic in this case.

haha I wrote the above forgetting about the big ol' sandbox

If you increase the empty space the "strange" result you were getting goes away, and behaves exactly as I describe above.

[docmerlin]

It would be more intuitive for the default to behave that way (either with increased space, or more realistically have the shapes move as if they are shapist.)

[vassudanagunta]

I mostly agree. But it is also important to consider that this should be an honest simulation. If you start tweaking it to get the results you want to see, it loses credibility. Another goal of this, I would think, is to get people thinking, not just consume this as proof or fact. The Internet is so partisan and dominated by junk because people are encouraged to confirm their own biases, rather than encouraged to think, to question, to understand.

It might be more useful for a section to be added at the end, with questions and exercises for the reader, like you see at ends of lessons in text books. For example:

1. Try setting the shapeism very high. Why does segregation decrease when you do that? Does this contradict what the earlier part of this page shows? What is really going on?
2. What happens when you increase the free area? Why?
3. When the shapes in this simulation move, they move randomly. But that's probably not what happens in the real world. How would the results change if the shapes moved only to spots that were better than their current spot? Or if they only moved to the best available spot according to their preferences?
4. What would happen if you had shapes with a variety of preferences (again, more like the real world), e.g. some triangles being very shapeist, some being less so, and some preferring diversity?

@ncase, I think your work is fantastic. I've been for many years toying on paper with an idea I call Illustrative Simulation, that would show many sorts of complex systems behavior such as yours, the Matthew Effect, Free Market dynamics, game theory such as the Prisoner's Dilemma, and even evolution. I'd love to collaborate with you!

[vassudanagunta]

@ncase, in my idea I envision being able to add more dimensions, or layers, to the scenario. For example, once the reader/student understands the forces behind segregation, the next step is to understand all the consequences beyond the apparent one (that people are segregated). What if the squares have more money and power than triangles? What if the quality of schools in a neighborhood are determined by the wealth and power of its residents? What if the prices of homes in an area where determined by the wealth of its residents (i.e. could triangles move to square neighborhoods even if they wanted to)?