There is some problem when a group contains only one agent (which can only happen in the homogenous-groups setup):
In this case the agent will be rep-agent every time and therefore never do admissibility calculations which can stall the run extremely if she is on a different theory than the one where the rest of the agents converge on.
Here is a picture from a run where that happened. At ~1200 ticks the run is still ongoing (agent 340 is the culprit):
Maybe we should force agents after being the rep-agent for a threshold number of times in a row to ignore the rep-agent restrictions?
What do you think? Any suggestions? :)
There is some problem when a group contains only one agent (which can only happen in the homogenous-groups setup): In this case the agent will be rep-agent every time and therefore never do admissibility calculations which can stall the run extremely if she is on a different theory than the one where the rest of the agents converge on.
Here is a picture from a run where that happened. At ~1200 ticks the run is still ongoing (agent 340 is the culprit):
Maybe we should force agents after being the rep-agent for a threshold number of times in a row to ignore the rep-agent restrictions? What do you think? Any suggestions? :)