Relationships #57
Description
I just stumbled across this tool from a post on linked in. It's a really cool idea, nicely implemented!
But I think as a first order model, it's coded slightly wrong. The arrows continue passing around the system, flipping on negatives, and modifying states. This means that the rate of change is purely a function of the number of clicks and the network structure.
You can this in the "depression" example - if you click the ^ on the "accepting mistakes" node once, and then let the simulation run, it equilibrates in a perfect no-depression state, because the arrows multiply due to the split at "doing things". If you give "anxiety" a ^ at the start, the opposite happens.
I think in general, the transfer functions should be a function of the input node state/size, not of the number of arrows.. What would make more sense is if the arrows were produced at a rate proportional to the size of each state, so the rate they are produced fluctuates as the node state value changes.
I think this would require two changes:
- Arrow generator: generates arrows at a constant rate, with the rate as a function of state size. You could also implement this as a random generator with a rate defined by the state size (like radioactive decay)
- Remove the arrows generated on click (but retain the state value modification)