One interesting elaboration I'd like to see investigated is including the directionality of communication. In this model, the information flow appears to be unidirectional, which is unrealistic, if we are modeling hierarchical structures.

Do nodes that give 'orders' to other nodes waste time? Should there be 'listeners' to organize and disseminate node-layer traffic?

I would imagine that including directionality, and partial directionality would give more relevant results per human organization.

Yes, the current model is undirectional. Superiors are only there to distribute information (essentially acting as routers) rather than acting as managers.

Maybe one could model leadership as getting information for decisions: a superior receives information about the state of the world from subordinates, and then makes decisions based on it that propagate down. There is degradation in the signal going both ways (the SNAFU principle). Performance would be measured by how little distortion could be achieved. In this simple model the flatter the organisation, the better. Here there would be a trade-off between overloading managers and having few levels. Most likely the answer is randomly connected people, making sure information always has plenty of shortcuts.

A more elaborate analysis would assume that the organisation meets problems of varying size and frequency, requiring different amounts of information. Since small local problems are more likely, it seems reasonable that it would promote more more connections in the periphery than the core.