Do agents perform their actions simultaneously? If yes, what's the point of shuffle_do()?

[PigeonZow]

Hi, I'm new to Mesa and had a question about how agents' actions are scheduled; Do agents perform their actions simultaneously, only using information available from the previous step? If yes, what is the point of shuffle_do() instead of do()? If not, is there a way to make actions happen "at the same time" (agents act based off of only the previous step's information)?

A naive google search led me to find the deprecated SimultaneousActivation from an older version of Mesa being replaced by do, but I'm still not completely sure.

My motivation for asking this is that some of the agents in my simulation make API calls and I'm trying to parallelize them.

Thanks!

[EwoutH]

Sorry for getting back so late.

No, agents don't act simultaneously by default. When you use self.agents.do("step"), they go one by one in whatever order they're in the AgentSet. Each agent sees the changes from agents that went before them.

Why shuffle_do() exists: It just randomizes who goes first. This matters because if Agent A always acts before Agent B, that could create biases in your model. Shuffling keeps things fair and adds some randomness to the order.

If you want actual simultaneous actions, you can use the two-phase approach that replaced SimultaneousActivation:

self.agents.do("step")      # Everyone decides what to do
self.agents.do("advance")   # Everyone does it

In your agent class:

def step(self):
    # Look around, decide what to do, but don't change anything yet
    self.planned_action = some_calculation()

def advance(self):
    # Now actually do the thing
    self.execute(self.planned_action)

For your API calls: If you're trying to parallelize API calls, that's a different beast. You'd probably want to look into Python's async/threading capabilities. The do() vs shuffle_do() thing is really about the logical order of agent activation in your simulation, not about actual parallel execution.

See our Migration guide for even more detail.

Does that answer your question?

[PigeonZow]

Yeah, that clears things up for me. Thanks a lot!